<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.4 20241031//EN" "JATS-journalpublishing1-4.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" article-type="research-article" dtd-version="1.4" xml:lang="en">
  <front>
    <journal-meta>
      <journal-id journal-id-type="publisher-id">airr</journal-id>
      <journal-title-group>
        <journal-title>Advances in Artificial Intelligence and Robotics Research</journal-title>
      </journal-title-group>
      <issn pub-type="epub">3143-3995</issn>
      <issn pub-type="ppub">3143-3987</issn>
      <publisher>
        <publisher-name>Scientific Research Publishing</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id pub-id-type="doi">10.4236/airr.2026.24012</article-id>
      <article-id pub-id-type="publisher-id">airr-155399</article-id>
      <article-categories>
        <subj-group>
          <subject>Article</subject>
        </subj-group>
        <subj-group>
          <subject>Computer Science</subject>
          <subject>Communications</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Machine Learning-Based Crop Recommendation Using Tuned Classifiers and Ensemble Methods</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Priya</surname>
            <given-names>Faiza Noushin</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Hasan</surname>
            <given-names>Mirza Md. Hasibul</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Akter</surname>
            <given-names>Mahfiza</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Badhon</surname>
            <given-names>Shahria</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0009-0007-0719-2200</contrib-id>
          <name name-style="western">
            <surname>Joya</surname>
            <given-names>Sraboni Ghosh</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
      </contrib-group>
      <aff id="aff1"><label>1</label> Department of Computer Science &amp; Engineering, City University, Dhaka, Bangladesh </aff>
      <author-notes>
        <fn fn-type="conflict" id="fn-conflict">
          <p>The authors declare no conflicts of interest regarding the publication of this paper.</p>
        </fn>
      </author-notes>
      <pub-date pub-type="epub">
        <day>01</day>
        <month>12</month>
        <year>2026</year>
      </pub-date>
      <pub-date pub-type="collection">
        <month>12</month>
        <year>2026</year>
      </pub-date>
      <volume>02</volume>
      <issue>04</issue>
      <fpage>209</fpage>
      <lpage>224</lpage>
      <history>
        <date date-type="received">
          <day>18</day>
          <month>09</month>
          <year>2026</year>
        </date>
        <date date-type="accepted">
          <day>15</day>
          <month>11</month>
          <year>2026</year>
        </date>
        <date date-type="published">
          <day>18</day>
          <month>11</month>
          <year>2026</year>
        </date>
      </history>
      <permissions>
        <copyright-statement>© 2026 by the authors and Scientific Research Publishing Inc.</copyright-statement>
        <copyright-year>2026</copyright-year>
        <license license-type="open-access">
          <license-p> This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ( <ext-link ext-link-type="uri" xlink:href="https://creativecommons.org/licenses/by/4.0/">https://creativecommons.org/licenses/by/4.0/</ext-link> ). </license-p>
        </license>
      </permissions>
      <self-uri content-type="doi" xlink:href="https://doi.org/10.4236/airr.2026.24012">https://doi.org/10.4236/airr.2026.24012</self-uri>
      <abstract>
        <p>The choice of a suitable crop for a given situation is determined by soil and various climatic parameters. For choosing these crops, we commonly use machine learning models, but combining multiple models, called ensembling, and checking whether an ensemble model can beat one well-tuned model is still an important question. In this paper, this is directly checked. We compared two publicly available crop recommendation datasets, the Ingle dataset and the Garg dataset, and selected the balanced Ingle dataset, containing 22 crops and 2200 records, as the sole dataset. Five models were trained under identical conditions: Logistic Regression, KNN, Random Forest, XGBoost, and Feedforward Neural Network. Next, we fine-tuned the hyperparameters of the top three models using Grid Search and incorporated them into the soft voting, hard voting, and stacking ensembles. To avoid data leakage, we used train, validation, and test sets with a clean split that ensured at no point did we have the same observations split between the three sets. The final results indicate that a well-calibrated Random Forest model obtained the same weighted F1 score (0.9954) as the three ensemble models. This demonstrates that if a single model is tuned correctly, introducing additional models will not improve performance on this dataset. This therefore provides a cost-effective, lightweight, and easy-to-install answer without running redundant models side by side for precision farming and for automatic crop selection applications.</p>
      </abstract>
      <kwd-group kwd-group-type="author-generated" xml:lang="en">
        <kwd>Crop Recommendation</kwd>
        <kwd>Machine Learning</kwd>
        <kwd>Ensemble Learning</kwd>
        <kwd>Random Forest</kwd>
        <kwd>Hyperparameter Tuning</kwd>
        <kwd>Precision Agriculture</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec1">
      <title>1. Introduction</title>
      <p>Selecting what to grow requires balancing soil chemistry (N, P, K, pH) and environment (humidity, temperature, rainfall) in complicated, non-linear ways, which is not very practical for traditional intuition. Prior work has found good accuracy with machine learning for crop selection using soil and climatic data [<xref ref-type="bibr" rid="B1">1</xref>]-[<xref ref-type="bibr" rid="B3">3</xref>]. However, many studies rely on default hyperparameters and a single train-test split, which gives little sense of reliability or generalizability. The task has been tackled by many classifiers, including boosted and tree-based models [<xref ref-type="bibr" rid="B4">4</xref>], deep neural networks [<xref ref-type="bibr" rid="B5">5</xref>], and comparisons of tree ensembles [<xref ref-type="bibr" rid="B6">6</xref>], yet few studies compare classifiers under systematic tuning. Ensembling is often used on the assumption that combining learners gives greater accuracy, rather than by comparing against a carefully optimized single learner [<xref ref-type="bibr" rid="B7">7</xref>] [<xref ref-type="bibr" rid="B8">8</xref>]. This leads to our main question: what happens when one systematically tunes the ensembled classifiers; does this lead to measurable improvements, or are the models simply reaching a limit set by the data? We explore this by systematically tuning each of the crop recommendation models and comparing its performance with ensembling. </p>
      <p>This problem is important not just for correctness, though. Unsuitable inputs use and practices can waste inputs, harm soil health, and constrain productivity [<xref ref-type="bibr" rid="B9">9</xref>] [<xref ref-type="bibr" rid="B10">10</xref>], especially for smallholder farmers who have limited flexibility to correct errors on their farms. So a good system will be conducive to good food security, sustainable use of land resources, and rationalization of the allocation of resources. With the increasing maturity of precision agriculture [<xref ref-type="bibr" rid="B11">11</xref>] [<xref ref-type="bibr" rid="B12">12</xref>], and the lack of high-end computing infrastructure in many locations [<xref ref-type="bibr" rid="B13">13</xref>], simplicity and deployability are just as important as accuracy, and the question is: Can a simple model perform a task as well as an ensemble? </p>
      <p>This study evaluates models on the Ingle benchmark dataset [<xref ref-type="bibr" rid="B14">14</xref>], consisting of 2200 samples across 7 features and 22 crop categories. Five models were tested and optimized using the Grid method and then ensembled in an appropriate way. The findings indicate that efficient crop recommendation can be achieved with the simple model “Random Forest” and that complexity is not always necessary to provide high levels of efficiency, since only linear limits exist to achieving accuracy. </p>
    </sec>
    <sec id="sec2">
      <title>2. Related Works</title>
      <p>It has been observed that ensemble-based methods are very effective for dealing with complex soil and weather data, and have replaced the simpler set of algorithms such as KNN, Decision Trees, and SVM for precision agriculture systems so far. These simpler models are outperformed by methods like Random Forest [<xref ref-type="bibr" rid="B15">15</xref>], boosting models like XGBoost and CatBoost [<xref ref-type="bibr" rid="B16">16</xref>], and neural networks [<xref ref-type="bibr" rid="B17">17</xref>]. For further improvement in accuracy, some recent studies stack or vote for more than one type of model: Ahmed <italic>et al</italic><italic>.</italic> added overfitting controls to the models [<xref ref-type="bibr" rid="B18">18</xref>]; Hasan <italic>et al</italic><italic>.</italic> used weighted averaging of a regression model and tree model [<xref ref-type="bibr" rid="B19">19</xref>]; Senapaty <italic>et al</italic><italic>.</italic> stacked SMOTE balancing and multiple models [<xref ref-type="bibr" rid="B20">20</xref>]. Other efforts are focused on practical deployment with basic hardware, for example, a low-cost IoT device that pairs soil sensors with a machine learning model for crop recommendation [<xref ref-type="bibr" rid="B13">13</xref>], and a hybrid GRU-SVR model for time-varying weather by Janani <italic>et al</italic><italic>.</italic> [<xref ref-type="bibr" rid="B21">21</xref>]. In conclusion, this body of work reveals that generally tree or boosting models perform better than traditional classifiers, and tuning plus data balancing (such as SMOTE) can increase prediction stability. <bold>Table 1</bold> summarizes the representative prior work in crop recommendation. </p>
      <p><bold>Table 1.</bold> Comparison of representative prior work in crop recommendation.</p>
      <table-wrap id="tbl1">
        <label>Table 1</label>
        <table>
          <tbody>
            <tr>
              <td>
                <bold>Paper</bold>
              </td>
              <td>
                <bold>Dataset</bold>
              </td>
              <td>
                <bold>Model</bold>
              </td>
              <td>
                <bold>Result</bold>
              </td>
              <td>
                <bold>Limit</bold>
                <bold>ations</bold>
              </td>
            </tr>
            <tr>
              <td>
                Ahmed
                <italic>et al</italic>
                <italic>.</italic>
                (2025) [
                <xref ref-type="bibr" rid="B18">18</xref>
                ]
              </td>
              <td>Two crop datasets with 28,242 records</td>
              <td>Complex 18-model ensemble</td>
              <td>Higher accuracy, lower overfitting</td>
              <td>High computational cost</td>
            </tr>
            <tr>
              <td>
                Hasan
                <italic>et al</italic>
                <italic>.</italic>
                (2023) [
                <xref ref-type="bibr" rid="B19">19</xref>
                ]
              </td>
              <td>Bangladesh farm data (1969-2021)</td>
              <td>KNN + RF + Ridge</td>
              <td>Better than SVR, NB, CatBoost</td>
              <td>Few crops; costs not checked</td>
            </tr>
            <tr>
              <td>
                Mahesh &amp; Soundrapandiyan (2024) [
                <xref ref-type="bibr" rid="B16">16</xref>
                ]
              </td>
              <td>Rainfall, pesticide, and temperature data</td>
              <td>CatBoost, LightGBM, XGBoost</td>
              <td>CatBoost worked best</td>
              <td>Needs many settings to be tuned</td>
            </tr>
            <tr>
              <td>
                Paithane (2023) [
                <xref ref-type="bibr" rid="B15">15</xref>
                ]
              </td>
              <td>Soil and crop data</td>
              <td>DT, SVM, LR, RF, NB, XGBoost</td>
              <td>Less overfitting; stable</td>
              <td>Not flexible; uneven data</td>
            </tr>
            <tr>
              <td>
                Ramzan
                <italic>et al</italic>
                <italic>.</italic>
                (2024) [
                <xref ref-type="bibr" rid="B17">17</xref>
                ]
              </td>
              <td>Live sensor + standard data</td>
              <td>Tuned Artificial Neural Network (ANN)</td>
              <td>99% accurate on both</td>
              <td>Needs data scaling; it is hard to explain</td>
            </tr>
            <tr>
              <td>
                Senapaty
                <italic>et al</italic>
                <italic>.</italic>
                (2024) [
                <xref ref-type="bibr" rid="B20">20</xref>
                ]
              </td>
              <td>Dataworld &amp; Odisha (246,000 records)</td>
              <td>SMOTE + SGDC</td>
              <td>100% Accurate; high F1 score</td>
              <td>Good accuracy but poor specificity</td>
            </tr>
            <tr>
              <td>
                Janani
                <italic>et al</italic>
                <italic>.</italic>
                (2026) [
                <xref ref-type="bibr" rid="B21">21</xref>
                ]
              </td>
              <td>Cauvery Delta (10 years’ data)</td>
              <td>Hybrid GRU + SVR (SKRDX)</td>
              <td>Predicts rapid weather changes</td>
              <td>Low accuracy when the weather changes rapidly</td>
            </tr>
          </tbody>
        </table>
      </table-wrap>
      <p>However, the existing studies are not without their drawbacks. While often high accuracy is often reported, minor performance problems such as low specificity for minority crops are neglected, as in the case of almost 100% accuracy but little specificity, which has been reported in Senapaty <italic>et al</italic><italic>.</italic> [<xref ref-type="bibr" rid="B20">20</xref>]. Many ensemble methods, such as the 18-model ensemble of Ahmed <italic>et al</italic><italic>.</italic> [<xref ref-type="bibr" rid="B18">18</xref>], are impractical for use due to expensive implementation in terms of the number of classifiers. Janani <italic>et al</italic><italic>.</italic>’s “hybrid” model compromised (relatively low) R2 quality because of the variation in climate [<xref ref-type="bibr" rid="B21">21</xref>]. Many studies also use just one train-test split instead of leakage-free cross-validations, so it is unclear if the models generalize, and the scaling of features in neural networks is a problem that makes their operation more complicated than tree-based models. Other work has addressed interpretability through explainable AI [<xref ref-type="bibr" rid="B22">22</xref>], evaluated a range of learning-based models [<xref ref-type="bibr" rid="B23">23</xref>], or treated recommendation as a multi-criteria problem [<xref ref-type="bibr" rid="B24">24</xref>] [<xref ref-type="bibr" rid="B25">25</xref>]. However, whether systematic tuning closes the gap between single classifiers and ensembles remains open. This study addresses that gap by investigating, through a systematic and direct comparison, whether a whole ensemble can outperform a well-tuned individual model.</p>
    </sec>
    <sec id="sec3">
      <title>3. Methodology</title>
      <sec id="sec3dot1">
        <title>3.1. Overview</title>
        <p>The proposed study is divided into the specified five stages, namely: dataset selection, data preprocessing, implementation of baseline models on five classifiers, hyperparameter tuning on the top three models of each group using the GridSearchCV approach, and ensemble construction. The overall workflow used in the process is shown in <xref ref-type="fig" rid="fig1">Figure 1</xref>.</p>
        <fig id="fig1">
          <label>Figure 1</label>
          <caption>
            <p>Proposed architecture for crop recommendation system.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId15.jpeg?20261118031448" />
        </fig>
        <p>All experiments were implemented using Python 3.13.15 with NumPy 2.1.3, Pandas 2.2.3, scikit-learn 1.6.1, and XGBoost 3.4.1.</p>
      </sec>
      <sec id="sec3dot2">
        <title>3.2. Dataset Description and Selection</title>
        <p>Two out of the publicly available two crop datasets were tested, which were acquired from Kaggle. The Ingle dataset [<xref ref-type="bibr" rid="B14">14</xref>] has 2200 rows evenly spread across 22 crop types, with seven features: Nitrogen (N), Phosphorus (P), Potassium (K), temperature, humidity, pH, and rainfall. The Garg dataset [<xref ref-type="bibr" rid="B26">26</xref>] has 6596 rows across 51 crop types with the same features, but the classes are not evenly distributed. The data was selected from the Ingle dataset due to the balance, which would mean that the class size would not affect the testing and tuning of different models used on it. A comparison of both datasets is given in <bold>Table 2</bold>. </p>
        <p><bold>Table 2.</bold> Comparison of the Ingle and Garg datasets.</p>
        <table-wrap id="tbl2">
          <label>Table 2</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Attribute</bold>
                </td>
                <td>
                  <bold>Ingle dataset</bold>
                </td>
                <td>
                  <bold>Garg dataset</bold>
                </td>
              </tr>
              <tr>
                <td>Samples</td>
                <td>2200</td>
                <td>6596</td>
              </tr>
              <tr>
                <td>Crop classes</td>
                <td>22</td>
                <td>51</td>
              </tr>
              <tr>
                <td>Features</td>
                <td>7 (N, P, K, temp., humidity, pH, rainfall)</td>
                <td>7 (N, P, K, temp., humidity, pH, rainfall)</td>
              </tr>
              <tr>
                <td>Class distribution</td>
                <td>Balanced (100 samples/class)</td>
                <td>Imbalanced</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
      </sec>
      <sec id="sec3dot3">
        <title>3.3. Preprocessing</title>
        <p>Only the training set (70%, 1540 samples) was used for normalization of all the numerical features. The validation set (20%, 440 samples) and test set (10%, 220 samples) were kept separate to avoid data leakage. The data were split using stratified sampling to preserve the class distribution, with a random seed of 42. Crop names were encoded as numerical labels. To ensure every dataset was transformed in exactly the same way, the data preparation and model training were combined into a single pipeline.</p>
      </sec>
      <sec id="sec3dot4">
        <title>3.4. Baseline Model Implementation</title>
        <p>In this proposed method, five machine learning models were used, which are Logistic Regression, K-Nearest Neighbors, Random Forest, XGBoost, and Feedforward Neural Network. They all underwent the same processing and were tested on the training set with the five-fold cross-validation method. </p>
        <p>1) Logistic Regression: Logistic Regression is an easy linear model that estimates the probability of a class as the result of a Sigmoid function applied to a linear combination of the input variables. It serves as a simple, interpretable baseline to compare against more complex algorithms. The model architecture is illustrated in <xref ref-type="fig" rid="fig2">Figure 2</xref>. </p>
        <fig id="fig2">
          <label>Figure 2</label>
          <caption>
            <p>Logistic regression architecture.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId16.jpeg?20261118031449" />
        </fig>
        <p>2) K-Nearest Neighbors: KNN is an instance-based classification approach that uses the majority class of the k nearest neighbors of the sample to classify it. It is suitable for this project because it does not require any assumption about how the data is distributed, and it captures local patterns of soil and climate data. The model architecture is illustrated in <xref ref-type="fig" rid="fig3">Figure 3</xref>. </p>
        <fig id="fig3">
          <label>Figure 3</label>
          <caption>
            <p>K-nearest neighbors architecture.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId17.jpeg?20261118031449" />
        </fig>
        <p>3) Random Forest: Random Forest is an ensemble of decision trees, which improves accuracy by using the majority voting principle for the final prediction. It has reduced overfitting compared to a single decision tree and is able to accommodate non-linear soil and environmental data. The model architecture is illustrated in <xref ref-type="fig" rid="fig4">Figure 4</xref>.</p>
        <fig id="fig4">
          <label>Figure 4</label>
          <caption>
            <p>Random forest architecture.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId18.jpeg?20261118031449" />
        </fig>
        <p>4) XGBoost: XGBoost is a gradient boosting algorithm in which boosted decision trees are created one by one in sequence with the aim of correcting the errors of their predecessors. It can also be used to regularize in order to reduce the risk of overfitting and extract more complex patterns from the data set. The model architecture is illustrated in <xref ref-type="fig" rid="fig5">Figure 5</xref>.</p>
        <fig id="fig5">
          <label>Figure 5</label>
          <caption>
            <p>XGBoost architecture.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId19.jpeg?20261118031449" />
        </fig>
        <p>5) Feedforward Neural Network: A Feedforward Neural Network is a network of connected neurons with weights and functions that process the input data. This is a framework that enables the network to learn non-linear features, which allows it to better capture non-linear relationships between soil and climate properties that can be weaker in simpler linear models. The model architecture is illustrated in <xref ref-type="fig" rid="fig6">Figure 6</xref><bold>.</bold></p>
        <fig id="fig6">
          <label>Figure 6</label>
          <caption>
            <p>Feedforward neural network architecture.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId20.jpeg?20261118031449" />
        </fig>
      </sec>
      <sec id="sec3dot5">
        <title>3.5. Hyperparameter Tuning</title>
        <p>To optimize the three models that were chosen, hyperparameter tuning was performed on them using GridSearchCV with five-fold cross-validation targeting the weighted F1 score. For Random Forest, tree count, max depth, and split criteria were explored. Tuning of XGBoost involved tuning and changing the number of trees, learning rate, depth, and subsample ratio. Adjusting hidden layer sizes, L2 regularization, and learning rates was performed for the neural network. For each model, the best configuration was selected by taking the one that achieved the best mean cross-validation score as shown in Equation (1).</p>
        <disp-formula id="FD1">
          <label>(1)</label>
          <mml:math display="inline">
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:msup>
                        <mml:mi>θ</mml:mi>
                        <mml:mo>*</mml:mo>
                      </mml:msup>
                      <mml:mo>=</mml:mo>
                      <mml:mo>
                      </mml:mo>
                      <mml:mi>arg</mml:mi>
                      <mml:msub>
                        <mml:mrow>
                          <mml:mi>max</mml:mi>
                        </mml:mrow>
                        <mml:mrow>
                          <mml:mi>θ</mml:mi>
                          <mml:mo>∈</mml:mo>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>Θ</mml:mi>
                          </mml:mstyle>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mfrac>
                            <mml:mn>1</mml:mn>
                            <mml:mstyle mathvariant="bold" mathsize="normal">
                              <mml:mi>K</mml:mi>
                            </mml:mstyle>
                          </mml:mfrac>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                      <mml:munderover>
                        <mml:mstyle mathsize="140%" displaystyle="true">
                          <mml:mo>∑</mml:mo>
                        </mml:mstyle>
                        <mml:mrow>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>k</mml:mi>
                          </mml:mstyle>
                          <mml:mo>=</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>K</mml:mi>
                        </mml:mstyle>
                      </mml:munderover>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>F</mml:mi>
                      </mml:mstyle>
                      <mml:msub>
                        <mml:mn>1</mml:mn>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>w</mml:mi>
                        </mml:mstyle>
                      </mml:msub>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mi>θ</mml:mi>
                          <mml:mo>,</mml:mo>
                          <mml:msub>
                            <mml:mstyle mathvariant="bold" mathsize="normal">
                              <mml:mi>D</mml:mi>
                            </mml:mstyle>
                            <mml:mstyle mathvariant="bold" mathsize="normal">
                              <mml:mi>k</mml:mi>
                            </mml:mstyle>
                          </mml:msub>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
      </sec>
      <sec id="sec3dot6">
        <title>3.6. Ensemble Construction</title>
        <p>Ensemble Construction Soft Voting, Hard Voting, and Stacking are three types of ensemble strategies constructed from the tuned base learners. Soft Voting takes the predicted probabilities from each base model, multiplies them by individual weights wi, and picks the class with the highest average probability, as defined in Equation (2). </p>
        <disp-formula id="FD2">
          <label>(2)</label>
          <mml:math>
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mtext>
                        </mml:mtext>
                        <mml:mover accent="true">
                          <mml:mtext>y</mml:mtext>
                          <mml:mo>^</mml:mo>
                        </mml:mover>
                        <mml:mrow>
                          <mml:mtext>soft</mml:mtext>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>=</mml:mo>
                      <mml:mtext>arg</mml:mtext>
                      <mml:munder>
                        <mml:mrow>
                          <mml:mi>max</mml:mi>
                        </mml:mrow>
                        <mml:mtext>c</mml:mtext>
                      </mml:munder>
                      <mml:munderover>
                        <mml:mstyle displaystyle="true" mathsize="140%">
                          <mml:mo>∑</mml:mo>
                        </mml:mstyle>
                        <mml:mrow>
                          <mml:mtext>i</mml:mtext>
                          <mml:mo>=</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                        <mml:mtext>n</mml:mtext>
                      </mml:munderover>
                      <mml:msub>
                        <mml:mtext>w</mml:mtext>
                        <mml:mtext>i</mml:mtext>
                      </mml:msub>
                      <mml:msub>
                        <mml:mtext>P</mml:mtext>
                        <mml:mtext>i</mml:mtext>
                      </mml:msub>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mtext>c</mml:mtext>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Hard Voting considers the final predicted class from each base model and chooses the class that receives the most votes, as shown in Equation (3). </p>
        <disp-formula id="FD3">
          <label>(3)</label>
          <mml:math display="inline">
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mtext>
                        </mml:mtext>
                        <mml:mover accent="true">
                          <mml:mtext>y</mml:mtext>
                          <mml:mo>^</mml:mo>
                        </mml:mover>
                        <mml:mrow>
                          <mml:mtext>hard</mml:mtext>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>=</mml:mo>
                      <mml:mtext>mode</mml:mtext>
                      <mml:mrow>
                        <mml:mo>{</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mtext>
                            </mml:mtext>
                            <mml:mover accent="true">
                              <mml:mtext>y</mml:mtext>
                              <mml:mo>^</mml:mo>
                            </mml:mover>
                            <mml:mn>1</mml:mn>
                          </mml:msub>
                          <mml:mo>,</mml:mo>
                          <mml:msub>
                            <mml:mtext>
                            </mml:mtext>
                            <mml:mover accent="true">
                              <mml:mtext>y</mml:mtext>
                              <mml:mo>^</mml:mo>
                            </mml:mover>
                            <mml:mn>2</mml:mn>
                          </mml:msub>
                          <mml:mo>,</mml:mo>
                          <mml:mo>⋅</mml:mo>
                          <mml:mo>⋅</mml:mo>
                          <mml:mo>⋅</mml:mo>
                          <mml:mo>,</mml:mo>
                          <mml:msub>
                            <mml:mtext>
                            </mml:mtext>
                            <mml:mover accent="true">
                              <mml:mtext>y</mml:mtext>
                              <mml:mo>^</mml:mo>
                            </mml:mover>
                            <mml:mtext>n</mml:mtext>
                          </mml:msub>
                        </mml:mrow>
                        <mml:mo>}</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Stacking uses a Logistic Regression model as a meta-learner on the predictions of the base learners. Instead of using a simple voting or averaging rule, it learns how to combine the base models’ outputs, as given in Equation (4). </p>
        <disp-formula id="FD4">
          <label>(4)</label>
          <mml:math>
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mtext>
                        </mml:mtext>
                        <mml:mover accent="true">
                          <mml:mtext>y</mml:mtext>
                          <mml:mo>^</mml:mo>
                        </mml:mover>
                        <mml:mrow>
                          <mml:mtext>stack</mml:mtext>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>=</mml:mo>
                      <mml:mtext>g</mml:mtext>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mtext>f</mml:mtext>
                            <mml:mn>1</mml:mn>
                          </mml:msub>
                          <mml:mrow>
                            <mml:mo>(</mml:mo>
                            <mml:mtext>X</mml:mtext>
                            <mml:mo>)</mml:mo>
                          </mml:mrow>
                          <mml:mo>,</mml:mo>
                          <mml:msub>
                            <mml:mtext>f</mml:mtext>
                            <mml:mn>2</mml:mn>
                          </mml:msub>
                          <mml:mrow>
                            <mml:mo>(</mml:mo>
                            <mml:mtext>X</mml:mtext>
                            <mml:mo>)</mml:mo>
                          </mml:mrow>
                          <mml:mo>,</mml:mo>
                          <mml:mo>⋅</mml:mo>
                          <mml:mo>⋅</mml:mo>
                          <mml:mo>⋅</mml:mo>
                          <mml:mo>,</mml:mo>
                          <mml:msub>
                            <mml:mtext>f</mml:mtext>
                            <mml:mtext>n</mml:mtext>
                          </mml:msub>
                          <mml:mrow>
                            <mml:mo>(</mml:mo>
                            <mml:mtext>X</mml:mtext>
                            <mml:mo>)</mml:mo>
                          </mml:mrow>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>After setting up these ensembles, grid search was used to tune them. This included trying out different types of voting configurations in the voting configuration and different regularization strengths for the Stacking meta-learner, and tuning the components of the meta-learner accordingly for each of the methods.</p>
      </sec>
      <sec id="sec3dot7">
        <title>3.7. Evaluation Metrics</title>
        <p>The performance of the model was assessed by accuracy, precision, recall, and weighted F1 score on both validation and test sets. The percentage of correct predictions for a class out of the total number of predictions made for a class is the precision measure, as shown in Equation (5). </p>
        <disp-formula id="FD5">
          <label>(5)</label>
          <mml:math>
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:mtext>Precision</mml:mtext>
                      <mml:mo>=</mml:mo>
                      <mml:mfrac>
                        <mml:mrow>
                          <mml:mtext>TP</mml:mtext>
                        </mml:mrow>
                        <mml:mrow>
                          <mml:mtext>TP</mml:mtext>
                          <mml:mo>+</mml:mo>
                          <mml:mtext>FP</mml:mtext>
                        </mml:mrow>
                      </mml:mfrac>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Recall is defined as the ratio of the number of samples that have been correctly predicted out of the total number of positive samples of a particular class, as defined in Equation (6). </p>
        <disp-formula id="FD6">
          <label>(6)</label>
          <mml:math>
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:mtext>Recall</mml:mtext>
                      <mml:mo>=</mml:mo>
                      <mml:mfrac>
                        <mml:mrow>
                          <mml:mtext>TP</mml:mtext>
                        </mml:mrow>
                        <mml:mrow>
                          <mml:mtext>TP</mml:mtext>
                          <mml:mo>+</mml:mo>
                          <mml:mtext>FN</mml:mtext>
                        </mml:mrow>
                      </mml:mfrac>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>The weighted F1 score is simply the harmonic average of the precision and recall values obtained per class, weighted by the number of samples in the class. It was adopted as the key criterion for model tuning and comparison in the presence of minor class imbalances that occurred during cross-validation, as given in Equation (7). </p>
        <disp-formula id="FD7">
          <label>(7)</label>
          <mml:math>
            <mml:mrow>
              <mml:mtable>
                <mml:mtr>
                  <mml:mtd>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mtext>F</mml:mtext>
                        <mml:mrow>
                          <mml:mn>1</mml:mn>
                          <mml:mtext>w</mml:mtext>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>=</mml:mo>
                      <mml:munderover>
                        <mml:mstyle mathsize="140%" displaystyle="true">
                          <mml:mo>∑</mml:mo>
                        </mml:mstyle>
                        <mml:mrow>
                          <mml:mtext>c</mml:mtext>
                          <mml:mo>=</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                        <mml:mtext>C</mml:mtext>
                      </mml:munderover>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mfrac>
                            <mml:mrow>
                              <mml:msub>
                                <mml:mtext>n</mml:mtext>
                                <mml:mtext>c</mml:mtext>
                              </mml:msub>
                            </mml:mrow>
                            <mml:mtext>N</mml:mtext>
                          </mml:mfrac>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                      <mml:mo>⋅</mml:mo>
                      <mml:mn>2</mml:mn>
                      <mml:mo>⋅</mml:mo>
                      <mml:mfrac>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mrow>
                              <mml:mtext>Precision</mml:mtext>
                            </mml:mrow>
                            <mml:mtext>c</mml:mtext>
                          </mml:msub>
                          <mml:mo>⋅</mml:mo>
                          <mml:msub>
                            <mml:mrow>
                              <mml:mtext>Recall</mml:mtext>
                            </mml:mrow>
                            <mml:mtext>c</mml:mtext>
                          </mml:msub>
                        </mml:mrow>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mrow>
                              <mml:mtext>Precision</mml:mtext>
                            </mml:mrow>
                            <mml:mtext>c</mml:mtext>
                          </mml:msub>
                          <mml:mo>+</mml:mo>
                          <mml:msub>
                            <mml:mrow>
                              <mml:mtext>Recall</mml:mtext>
                            </mml:mrow>
                            <mml:mtext>c</mml:mtext>
                          </mml:msub>
                        </mml:mrow>
                      </mml:mfrac>
                    </mml:mrow>
                  </mml:mtd>
                </mml:mtr>
              </mml:mtable>
            </mml:mrow>
          </mml:math>
        </disp-formula>
      </sec>
    </sec>
    <sec id="sec4">
      <title>4. Results and Discussion</title>
      <sec id="sec4dot1">
        <title>4.1. Dataset Selection Outcome</title>
        <p>From the two sets of data considered, the “balanced” 2200-sample Ingle dataset (22 classes) was selected for the remaining experiments. All results from here on are based on this dataset. This is a comparison of the baseline models.</p>
      </sec>
      <sec id="sec4dot2">
        <title>4.2. Baseline Model Comparison</title>
        <p><bold>Table 3.</bold>Baseline model performance on the Ingle dataset.</p>
        <table-wrap id="tbl3">
          <label>Table 3</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Model</bold>
                </td>
                <td>
                  <bold>Accuracy</bold>
                </td>
                <td>
                  <bold>Precision</bold>
                </td>
                <td>
                  <bold>Recall</bold>
                </td>
                <td>
                  <bold>F1-score</bold>
                </td>
              </tr>
              <tr>
                <td>Random Forest</td>
                <td>0.9955</td>
                <td>0.9957</td>
                <td>0.9955</td>
                <td>0.9955</td>
              </tr>
              <tr>
                <td>XGBoost</td>
                <td>0.9932</td>
                <td>0.9935</td>
                <td>0.9932</td>
                <td>0.9931</td>
              </tr>
              <tr>
                <td>FFN</td>
                <td>0.9909</td>
                <td>0.9915</td>
                <td>0.9909</td>
                <td>0.9909</td>
              </tr>
              <tr>
                <td>KNN</td>
                <td>0.9795</td>
                <td>0.9804</td>
                <td>0.9795</td>
                <td>0.9793</td>
              </tr>
              <tr>
                <td>Logistic Regression</td>
                <td>0.9727</td>
                <td>0.9740</td>
                <td>0.9727</td>
                <td>0.9725</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <fig id="fig7">
          <label>Figure 7</label>
          <caption>
            <p>Baseline model performance comparison across four metrics.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId35.jpeg?20261118031450" />
        </fig>
        <p>The following models were tested as baseline models: Logistic Regression, K-Nearest Neighbors, Random Forest, XGBoost, and a Feedforward Neural Network. All models were preprocessed in the same way and evaluated using five-fold stratified cross-validation on the training set. The 1540 training samples were divided into five stratified folds, with four folds used for training and the remaining fold used for validation in each iteration. This process was repeated five times so that each fold served as the validation fold once. The reported baseline performance represents the average performance across the five folds. The separate validation set (440 samples) and test set (220 samples) were not used during this cross-validation. The performance metrics are shown in <bold>Table 3</bold> and <xref ref-type="fig" rid="fig7">Figure 7</xref>. </p>
        <p>Random Forest and XGBoost achieved the best results with an average weighted F1-score of 0.9955 and 0.9931, respectively. Logistic Regression had the highest standard deviation across the folds (std = 0.0099), possibly due to a lack of good performance with non-linear patterns. Based on these results, Random Forest, XGBoost, and the Neural Network were chosen for hyperparameter tuning. </p>
      </sec>
      <sec id="sec4dot3">
        <title>4.3. The Hyperparameter Tuning</title>
        <p>The three highest baselines were selected and optimized further. GridSearchCV with five-fold stratified cross-validation was used to search for the best hyperparameters by maximizing the weighted F1 score. The search was performed using the training set only. The complete hyperparameter search spaces are presented in <bold>Table 4</bold>, while <bold>Table 5</bold> lists the best configurations identified by GridSearchCV. </p>
        <p><bold>Table 4.</bold> Complete hyperparameter search spaces.</p>
        <table-wrap id="tbl4">
          <label>Table 4</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Model</bold>
                </td>
                <td>
                  <bold>Hyperparameter</bold>
                </td>
                <td>
                  <bold>Search values</bold>
                </td>
              </tr>
              <tr>
                <td rowspan="4">Random Forest</td>
                <td>n_estimators</td>
                <td>100, 200</td>
              </tr>
              <tr>
                <td>max_depth</td>
                <td>None, 10, 20</td>
              </tr>
              <tr>
                <td>min_samples_split</td>
                <td>2, 5</td>
              </tr>
              <tr>
                <td>min_samples_leaf</td>
                <td>1, 2</td>
              </tr>
              <tr>
                <td rowspan="4">XGBoost</td>
                <td>n_estimators</td>
                <td>100, 200</td>
              </tr>
              <tr>
                <td>max_depth</td>
                <td>3, 6, 9</td>
              </tr>
              <tr>
                <td>learning_rate</td>
                <td>0.01, 0.1, 0.3</td>
              </tr>
              <tr>
                <td>subsample</td>
                <td>0.8, 1.0</td>
              </tr>
              <tr>
                <td rowspan="3">FFN (MLP)</td>
                <td>hidden_layer_sizes</td>
                <td>(64), (128, 64), (128, 64, 32)</td>
              </tr>
              <tr>
                <td>alpha</td>
                <td>0.0001, 0.001, 0.01</td>
              </tr>
              <tr>
                <td>learning_rate_init</td>
                <td>0.001, 0.01</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <p><bold>Table 5.</bold>Best hyperparameter configurations identified via GridSearchCV.</p>
        <table-wrap id="tbl5">
          <label>Table 5</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Model</bold>
                </td>
                <td>
                  <bold>Best configuration</bold>
                </td>
                <td>
                  <bold>CV F1</bold>
                </td>
              </tr>
              <tr>
                <td>Random Forest</td>
                <td>n_estimators = 100, max_depth = None, min_samples_split = 5, min_samples_leaf = 1</td>
                <td>0.9954</td>
              </tr>
              <tr>
                <td>XGBoost</td>
                <td>n_estimators = 200, max_depth = 6, learning_rate = 0.3, subsample = 1.0</td>
                <td>0.9896</td>
              </tr>
              <tr>
                <td>FFN (MLP)</td>
                <td>hidden_layer_sizes = (64), alpha = 0.01, learning_rate_init = 0.01</td>
                <td>0.9870</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <p>The FFN used ReLU activation, a maximum of 500 training iterations, and a random state of 42. <bold>Table 6</bold> and <xref ref-type="fig" rid="fig8">Figure 8</xref> illustrate the effect of the tuning. These present the cross-validation weighted F1 scores of each model before and after hyperparameter tuning. </p>
        <p><bold>Table 6.</bold>Model performance before vs. after hyperparameter tuning.</p>
        <table-wrap id="tbl6">
          <label>Table 6</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Model</bold>
                </td>
                <td>
                  <bold>CV F1 (before)</bold>
                </td>
                <td>
                  <bold>CV F1 (after)</bold>
                </td>
                <td>
                  <bold>Change</bold>
                </td>
              </tr>
              <tr>
                <td>Random Forest</td>
                <td>0.9950</td>
                <td>0.9954</td>
                <td>+0.0004</td>
              </tr>
              <tr>
                <td>XGBoost</td>
                <td>0.9927</td>
                <td>0.9896</td>
                <td>−0.0031</td>
              </tr>
              <tr>
                <td>FFN (MLP)</td>
                <td>0.9859</td>
                <td>0.9870</td>
                <td>+0.0011</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <fig id="fig8">
          <label>Figure 8</label>
          <caption>
            <p>Before vs. after hyperparameter tuning comparison.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId36.jpeg?20261118031450" />
        </fig>
        <p>There was little difference on this dataset with tuning. Random Forest improved by 0.0004 F1 score and Neural Network by 0.0011 F1 score, and XGBoost decreased by 0.0031 F1 score after tuning. It is a small decrease that indicates the use of the default hyperparameters of XGBoost was near the optimum. Overall, the results show that all three models were already performing near the upper range observed on this dataset before tuning. </p>
      </sec>
      <sec id="sec4dot4">
        <title>4.4. Ensemble Learning</title>
        <p>The following three ensemble methods were built: Soft Voting, Hard Voting, and Stacking (with Logistic Regression as the meta-learner). For stacking, we employed a careful approach to prevent data leakage during meta-learner training. The base models (RF, XGBoost, and FNN) were trained using 5-fold cross-validation, where predictions on held-out folds in each iteration served as training data for the Logistic Regression meta-learner. This out-of-fold approach ensured that the meta-learner was trained using predictions generated from held-out samples rather than predictions from samples used to train the corresponding base models. Once trained, the meta-learner combined base model predictions on the final test set. Before and after the ensemble tuning, the results of the performance are shown in <bold>Table 7</bold> and <xref ref-type="fig" rid="fig9">Figure 9</xref>. In this configuration, the term “before tuning” includes those ensembles where all base models had equal weights. “After tuning” means the voting weights and the parameters of the meta-learner were tuned with the grid search.</p>
        <p>For Soft Voting, the weight combinations [1, 1, 1], [2, 1, 1], [1, 2, 1], [1, 1, 2], [2, 2, 1], and [3, 2, 1] were evaluated using GridSearchCV. For Stacking, the Logistic Regression metalearner was tuned using C values of 0.1, 1.0, and 10.0, and the passthrough option was evaluated as False and True. Both searches used 5-fold cross-validation and the weighted F1 score as the scoring metric. </p>
        <p><bold>Table 7.</bold>Ensemble performance before vs. after ensemble-level tuning.</p>
        <table-wrap id="tbl7">
          <label>Table 7</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Ensemble</bold>
                </td>
                <td>
                  <bold>CV F1 (before)</bold>
                </td>
                <td>
                  <bold>CV F1 (after)</bold>
                </td>
                <td>
                  <bold>Change</bold>
                </td>
              </tr>
              <tr>
                <td>Soft Voting</td>
                <td>0.9932</td>
                <td>0.9954</td>
                <td>+0.0022</td>
              </tr>
              <tr>
                <td>Hard Voting</td>
                <td>0.9932</td>
                <td>0.9954</td>
                <td>+0.0022</td>
              </tr>
              <tr>
                <td>Stacking</td>
                <td>0.9932</td>
                <td>0.9948</td>
                <td>+0.0016</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <fig id="fig9">
          <label>Figure 9</label>
          <caption>
            <p>Before vs. after ensemble tuning comparison.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId37.jpeg?20261118031451" />
        </fig>
        <p><bold>Table 8.</bold> Final comparison of tuned single models vs. tuned ensembles.</p>
        <table-wrap id="tbl8">
          <label>Table 8</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Rank</bold>
                </td>
                <td>
                  <italic>Model</italic>
                </td>
                <td>
                  <italic>CV weighted</italic>
                  F1
                </td>
              </tr>
              <tr>
                <td>1</td>
                <td>Random Forest (tuned)</td>
                <td>0.9954</td>
              </tr>
              <tr>
                <td>1</td>
                <td>Voting Soft (2, 2, 1)</td>
                <td>0.9954</td>
              </tr>
              <tr>
                <td>1</td>
                <td>Voting Hard (1, 1, 1)</td>
                <td>0.9954</td>
              </tr>
              <tr>
                <td>4</td>
                <td>Stacking (C = 1.0)</td>
                <td>0.9948</td>
              </tr>
              <tr>
                <td>5</td>
                <td>XGBoost (tuned)</td>
                <td>0.9896</td>
              </tr>
              <tr>
                <td>6</td>
                <td>FFN/MLP (tuned)</td>
                <td>0.9870</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <p>In addition, ensemble tuning was more beneficial than performance with single model tuning. Random Forest and XGBoost outperformed the neural network—Soft Voting improved by 0.0022 when these two models received double the weight of the neural network, and equal weights were the best for Hard Voting. The empowerment of both voting schemes was 0.9954. Stacking improved by 0.0016 after tuning its regularization to obtain a value of 0.9948. There were not many opportunities for improvement with the individual models, but if these were all combined with the optimum weighting, a fairly significant improvement was achieved, whose parameters were definitely measurable. In <bold>Table 8</bold>, and in <xref ref-type="fig" rid="fig10">Figure 10</xref>, the final rankings of all six tuned models are given. This includes the three individual base models along with three ensemble strategies (all ordered by the cross-validation weighted F1). </p>
        <fig id="fig10">
          <label>Figure 10</label>
          <caption>
            <p>Final comparison of tuned single models vs. tuned ensembles.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/2970159-rId38.jpeg?20261118031451" />
        </fig>
        <p>After tuning, the Random Forest and the voting ensembles achieved the highest cross-validation weighted F1 score of 0.9954. This finding reveals that the combination of multiple models does not necessarily outperform a single well-tuned model. With stacking, the outcome was similar, as the models at the bottom level predicted similarly and there was little information for the meta-learner to exploit. There is a trade-off between model stability, accuracy, and simplicity: tuned voting ensembles are useful when even more stable predictions are sought after, but Random Forest is the model of choice when it is needed to be not only accurate but also simple. With correct modelling, however, the gain from ensembling is almost negligible. </p>
        <p>To further evaluate the generalization performance of the final selected models, each model was evaluated once on the untouched test set containing 220 samples. The test set was not used during model or ensemble selection. The final test-set performance is reported in <bold>Table 9</bold>.</p>
        <p><bold>Table 9.</bold>Final performance on the untouched test set.</p>
        <table-wrap id="tbl9">
          <label>Table 9</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Model</bold>
                </td>
                <td>
                  <bold>Accuracy</bold>
                </td>
                <td>
                  <bold>Precision</bold>
                </td>
                <td>
                  <bold>Recall</bold>
                </td>
                <td>
                  <bold>Weighted F1</bold>
                </td>
              </tr>
              <tr>
                <td>RF (tuned)</td>
                <td>1.0000</td>
                <td>1.0000</td>
                <td>1.0000</td>
                <td>1.0000</td>
              </tr>
              <tr>
                <td>XGB (tuned)</td>
                <td>0.9864</td>
                <td>0.9876</td>
                <td>0.9864</td>
                <td>0.9863</td>
              </tr>
              <tr>
                <td>FFN (tuned)</td>
                <td>0.9909</td>
                <td>0.9917</td>
                <td>0.9909</td>
                <td>0.9909</td>
              </tr>
              <tr>
                <td>Soft Vote</td>
                <td>0.9955</td>
                <td>0.9959</td>
                <td>0.9955</td>
                <td>0.9954</td>
              </tr>
              <tr>
                <td>Hard Vote</td>
                <td>1.0000</td>
                <td>1.0000</td>
                <td>1.0000</td>
                <td>1.0000</td>
              </tr>
              <tr>
                <td>Stacking</td>
                <td>1.0000</td>
                <td>1.0000</td>
                <td>1.0000</td>
                <td>1.0000</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <p>On the untouched test set, Random Forest, Hard Voting, and Stacking achieved 1.0000 across all reported metrics. Soft Voting achieved a weighted F1 score of 0.9954, while FFN and XGBoost achieved weighted F1 scores of 0.9909 and 0.9863, respectively. These results provide an independent evaluation of the final selected models after the model and ensemble selection process. </p>
      </sec>
    </sec>
    <sec id="sec5">
      <title>5. Conclusions</title>
      <p>This study compared two publicly available crop recommendation datasets from Kaggle and selected the Ingle dataset as the sole dataset for model evaluation, containing 2200 samples across 22 classes as a balanced benchmark. Five baseline models were tested, and three of the highest-performing (Random Forest, XGBoost, Feedforward NN) were tuned using grid search to create these ensembles: Soft Voting, Hard Voting, and Stacking. There was no gain in combining models, as all the well-tuned ensembles achieved the same best cross-validation weighted F1 Score as the tuned Random Forest (0.9954). The main contribution of this work is a systematic comparison of single classifiers against ensembles, demonstrating that a well-tuned Random Forest can match complex ensembles on the balanced Ingle dataset used as the evaluation benchmark. </p>
      <p>Future Work: These are the areas to be undertaken as future work based on our results: </p>
      <p>Final Model Selection: Considering both the tuned Random Forest and the tuned Voting ensemble, as well as the equal performance, computational load, and ease of deployment of these models, select the one to be used.Extended Model Benchmarking: Try out some other algorithms such as Support Vector Machines, LightGBM, and CatBoost, and extend the pipeline comparisons. Real World Data Validation: Apply the selected model to larger datasets of field-collected data to test robustness in real-world agricultural conditions and not clean data. Development: Create a simple, lightweight application for farmers or local agricultural workers to use easily with the selected model. </p>
    </sec>
    <sec id="sec6">
      <title>Data Availability</title>
      <p>The datasets analyzed during the current study are publicly available in the Kaggle repository. The primary dataset used for model development is available at <ext-link ext-link-type="uri" xlink:href="https://www.kaggle.com/datasets/atharvaingle/crop-recommendation-dataset">https://www.kaggle.com/datasets/atharvaingle/crop-recommendation-dataset</ext-link> [<xref ref-type="bibr" rid="B14">14</xref>]. The comparison dataset is available at <ext-link ext-link-type="uri" xlink:href="https://www.kaggle.com/datasets/sashankgarg23/crop-dataset">https://www.kaggle.com/datasets/sashankgarg23/crop-dataset</ext-link> [<xref ref-type="bibr" rid="B26">26</xref>].</p>
    </sec>
    <sec id="sec7">
      <title>Acknowledgements</title>
      <p>The authors thank the Department of Computer Science &amp; Engineering, City University, Dhaka, Bangladesh, and gratefully acknowledge the supervision of Sraboni Ghosh Joya throughout this study. </p>
    </sec>
    <sec id="sec8">
      <title>Author Contributions</title>
      <p>Conceptualization, S.G.J., F.N.P., M.M.H.H., M.A. and S.B.; methodology, F.N.P.; software, F.N.P.; validation, F.N.P. and M.M.H.H.; formal analysis, M.M.H.H. and S.B.; investigation, M.A.; resources, S.B. and M.A.; data curation; writing original draft preparation, F.N.P., M.A., S.B. and M.M.H.H.; writing review and editing, S.G.J. and F.N.P.; visualization, F.N.P.; supervision, S.G.J.; project administration, S.G.J.; funding acquisition, F.N.P., M.M.H.H., M.A., S.B. and S.G.J. All authors have read and agreed to the published version of the manuscript.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <title>References</title>
      <ref id="B1">
        <label>1.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Gosai, D., Raval, C., Nayak, R., Jayswal, H. and Patel, A. (2021) Crop Recommendation System Using Machine Learning. <italic>International</italic><italic>Journal</italic><italic>of</italic><italic>Scientific</italic><italic>Research</italic><italic>in</italic><italic>Computer</italic><italic>Science</italic>, <italic>Engineering</italic><italic>and</italic><italic>Information</italic><italic>Technology</italic>, 7, 558-569. https://doi.org/10.32628/cseit2173129 <pub-id pub-id-type="doi">10.32628/cseit2173129</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.32628/cseit2173129">https://doi.org/10.32628/cseit2173129</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Gosai, D.</string-name>
              <string-name>Raval, C.</string-name>
              <string-name>Nayak, R.</string-name>
              <string-name>Jayswal, H.</string-name>
              <string-name>Patel, A.</string-name>
              <string-name>Science, E</string-name>
            </person-group>
            <year>2021</year>
            <article-title>Crop Recommendation System Using Machine Learning</article-title>
            <source>International Journal of Scientific Research in Computer Science</source>
            <volume>7</volume>
            <pub-id pub-id-type="doi">10.32628/cseit2173129</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B2">
        <label>2.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Apat, S.K., Mishra, J., Raju, K.S. and Padhy, N. (2023) An Artificial Intelligence-Based Crop Recommendation System Using Machine Learning. <italic>Journal of Scientific &amp; Industrial Research</italic>, 82, 558-567.</mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Apat, S.K.</string-name>
              <string-name>Mishra, J.</string-name>
              <string-name>Raju, K.S.</string-name>
              <string-name>Padhy, N.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>An Artificial Intelligence-Based Crop Recommendation System Using Machine Learning</article-title>
            <source>Journal of Scientific &amp; Industrial Research</source>
            <volume>82</volume>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B3">
        <label>3.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Afzal, H., Amjad, M., Raza, A., Munir, K., Villar, S.G., Lopez, L.A.D., <italic>et al</italic>. (2025) Incorporating Soil Information with Machine Learning for Crop Recommendation to Improve Agricultural Output. <italic>Scientific Reports</italic>, 15, Article No. 8560. https://doi.org/10.1038/s41598-025-88676-z <pub-id pub-id-type="doi">10.1038/s41598-025-88676-z</pub-id><pub-id pub-id-type="pmid">40074741</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1038/s41598-025-88676-z">https://doi.org/10.1038/s41598-025-88676-z</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Afzal, H.</string-name>
              <string-name>Amjad, M.</string-name>
              <string-name>Raza, A.</string-name>
              <string-name>Munir, K.</string-name>
              <string-name>Villar, S.G.</string-name>
              <string-name>Lopez, L.A.D.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>Incorporating Soil Information with Machine Learning for Crop Recommendation to Improve Agricultural Output</article-title>
            <source>Scientific Reports</source>
            <volume>15</volume>
            <elocation-id>No</elocation-id>
            <pub-id pub-id-type="doi">10.1038/s41598-025-88676-z</pub-id>
            <pub-id pub-id-type="pmid">40074741</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B4">
        <label>4.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Shams, M.Y., Gamel, S.A. and Talaat, F.M. (2024) Enhancing Crop Recommendation Systems with Explainable Artificial Intelligence: A Study on Agricultural Decision-Making. <italic>Neural Com</italic><italic>puting and Applications</italic>, 36, 5695-5714. https://doi.org/10.1007/s00521-023-09391-2 <pub-id pub-id-type="doi">10.1007/s00521-023-09391-2</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1007/s00521-023-09391-2">https://doi.org/10.1007/s00521-023-09391-2</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Shams, M.Y.</string-name>
              <string-name>Gamel, S.A.</string-name>
              <string-name>Talaat, F.M.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Enhancing Crop Recommendation Systems with Explainable Artificial Intelligence: A Study on Agricultural Decision-Making</article-title>
            <source>Neural Computing and Applications</source>
            <volume>36</volume>
            <pub-id pub-id-type="doi">10.1007/s00521-023-09391-2</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B5">
        <label>5.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Venkateswara, S.M. and Padmanaban, J. (2025) Interpretable Deep Learning Models for Independent Fertilizer and Crop Recommendation. <italic>Scientific Reports</italic>, 15, Article No. 41721. https://doi.org/10.1038/s41598-025-26910-4 <pub-id pub-id-type="doi">10.1038/s41598-025-26910-4</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1038/s41598-025-26910-4">https://doi.org/10.1038/s41598-025-26910-4</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Venkateswara, S.M.</string-name>
              <string-name>Padmanaban, J.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>Interpretable Deep Learning Models for Independent Fertilizer and Crop Recommendation</article-title>
            <source>Scientific Reports</source>
            <volume>15</volume>
            <elocation-id>No</elocation-id>
            <pub-id pub-id-type="doi">10.1038/s41598-025-26910-4</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B6">
        <label>6.</label>
        <citation-alternatives>
          <mixed-citation publication-type="book">Yadav, S., Singh, H.K., Garg, A. and Khan, M.M. (2026) Comparative Analysis of Machine Learning Approaches for Crop Recommendation in Sustainable Agriculture in India. In: <italic>Atlantis Highlights in Intelligent Systems</italic>, Atlantis Press International BV, 578-586. https://doi.org/10.2991/978-94-6239-693-7_57 <pub-id pub-id-type="doi">10.2991/978-94-6239-693-7_57</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.2991/978-94-6239-693-7_57">https://doi.org/10.2991/978-94-6239-693-7_57</ext-link></mixed-citation>
          <element-citation publication-type="book">
            <person-group person-group-type="author">
              <string-name>Yadav, S.</string-name>
              <string-name>Singh, H.K.</string-name>
              <string-name>Garg, A.</string-name>
              <string-name>Khan, M.M.</string-name>
              <string-name>Systems, A</string-name>
            </person-group>
            <year>2026</year>
            <article-title>Comparative Analysis of Machine Learning Approaches for Crop Recommendation in Sustainable Agriculture in India</article-title>
            <source>In: Atlantis Highlights in Intelligent Systems</source>
            <volume>578</volume>
            <pub-id pub-id-type="doi">10.2991/978-94-6239-693-7_57</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B7">
        <label>7.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Acharya, N., Khatiwada, P., Pandey, R., Niroula, S. and Chapagain, P. (2024) Crop Recommendation System Using Machine Learning: A Comparative Study. <italic>International Journal on Engine</italic><italic>ering Technology</italic>, 1, 302-311. https://doi.org/10.3126/injet.v1i2.66708 <pub-id pub-id-type="doi">10.3126/injet.v1i2.66708</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3126/injet.v1i2.66708">https://doi.org/10.3126/injet.v1i2.66708</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Acharya, N.</string-name>
              <string-name>Khatiwada, P.</string-name>
              <string-name>Pandey, R.</string-name>
              <string-name>Niroula, S.</string-name>
              <string-name>Chapagain, P.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Crop Recommendation System Using Machine Learning: A Comparative Study</article-title>
            <source>International Journal on Engineering Technology</source>
            <volume>1</volume>
            <pub-id pub-id-type="doi">10.3126/injet.v1i2.66708</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B8">
        <label>8.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Rinayat, B.S. and Dharaskar, B. (2025) A Robust Ensemble Approach for Crop Recommendation Based on Environmental Parameters. <italic>EPJ Web of Conferences</italic>, 328, Article 01024. https://doi.org/10.1051/epjconf/202532801024 <pub-id pub-id-type="doi">10.1051/epjconf/202532801024</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1051/epjconf/202532801024">https://doi.org/10.1051/epjconf/202532801024</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Rinayat, B.S.</string-name>
              <string-name>Dharaskar, B.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>A Robust Ensemble Approach for Crop Recommendation Based on Environmental Parameters</article-title>
            <source>EPJ Web of Conferences</source>
            <volume>328</volume>
            <elocation-id>01024</elocation-id>
            <pub-id pub-id-type="doi">10.1051/epjconf/202532801024</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B9">
        <label>9.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Khaliq, A., Khan, A., Jan, S., Umair, M., Gulshair, A., Ali, A., <italic>et al</italic>. (2025) AI-Driven Smart Agriculture: An Integrated Approach for Soil Analysis, Irrigation, and Crop-Fertilizer Recommendations. <italic>IEEE Access</italic>, 13, 141124-141138. https://doi.org/10.1109/access.2025.3594162 <pub-id pub-id-type="doi">10.1109/access.2025.3594162</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1109/access.2025.3594162">https://doi.org/10.1109/access.2025.3594162</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Khaliq, A.</string-name>
              <string-name>Khan, A.</string-name>
              <string-name>Jan, S.</string-name>
              <string-name>Umair, M.</string-name>
              <string-name>Gulshair, A.</string-name>
              <string-name>Ali, A.</string-name>
              <string-name>Analysis, I</string-name>
            </person-group>
            <year>2025</year>
            <article-title>AI-Driven Smart Agriculture: An Integrated Approach for Soil Analysis, Irrigation, and Crop-Fertilizer Recommendations</article-title>
            <source>IEEE Access</source>
            <volume>13</volume>
            <pub-id pub-id-type="doi">10.1109/access.2025.3594162</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B10">
        <label>10.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Goap, A., Sharma, D., Shukla, A.K. and Krishna, C.R. (2018) An IoT Based Smart Irrigation Management System Using Machine Learning and Open Source Technologies. <italic>Computers</italic><italic>and</italic><italic>Electronics</italic><italic>in</italic><italic>Agriculture</italic>, 155, 41-49. https://doi.org/10.1016/j.compag.2018.09.040 <pub-id pub-id-type="doi">10.1016/j.compag.2018.09.040</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.compag.2018.09.040">https://doi.org/10.1016/j.compag.2018.09.040</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Goap, A.</string-name>
              <string-name>Sharma, D.</string-name>
              <string-name>Shukla, A.K.</string-name>
              <string-name>Krishna, C.R.</string-name>
            </person-group>
            <year>2018</year>
            <article-title>An IoT Based Smart Irrigation Management System Using Machine Learning and Open Source Technologies</article-title>
            <source>Computers and Electronics in Agriculture</source>
            <volume>155</volume>
            <pub-id pub-id-type="doi">10.1016/j.compag.2018.09.040</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B11">
        <label>11.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Bakthavatchalam, K., Karthik, B., Thiruvengadam, V., Muthal, S., Jose, D., Kotecha, K., <italic>et al</italic>. (2022) IoT Framework for Measurement and Precision Agriculture: Predicting the Crop Using Machine Learning Algorithms. <italic>Technologies</italic>, 10, Article 13. https://doi.org/10.3390/technologies10010013 <pub-id pub-id-type="doi">10.3390/technologies10010013</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3390/technologies10010013">https://doi.org/10.3390/technologies10010013</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Bakthavatchalam, K.</string-name>
              <string-name>Karthik, B.</string-name>
              <string-name>Thiruvengadam, V.</string-name>
              <string-name>Muthal, S.</string-name>
              <string-name>Jose, D.</string-name>
              <string-name>Kotecha, K.</string-name>
            </person-group>
            <year>2022</year>
            <article-title>IoT Framework for Measurement and Precision Agriculture: Predicting the Crop Using Machine Learning Algorithms</article-title>
            <source>Technologies</source>
            <volume>10</volume>
            <elocation-id>13</elocation-id>
            <pub-id pub-id-type="doi">10.3390/technologies10010013</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B12">
        <label>12.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Thilakarathne, N.N., Bakar, M.S.A., Abas, P.E. and Yassin, H. (2022) A Cloud Enabled Crop Recommendation Platform for Machine Learning-Driven Precision Farming. <italic>Sensors</italic>, 22, Article 6299. https://doi.org/10.3390/s22166299 <pub-id pub-id-type="doi">10.3390/s22166299</pub-id><pub-id pub-id-type="pmid">36016060</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3390/s22166299">https://doi.org/10.3390/s22166299</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Thilakarathne, N.N.</string-name>
              <string-name>Bakar, M.S.A.</string-name>
              <string-name>Abas, P.E.</string-name>
              <string-name>Yassin, H.</string-name>
            </person-group>
            <year>2022</year>
            <article-title>A Cloud Enabled Crop Recommendation Platform for Machine Learning-Driven Precision Farming</article-title>
            <source>Sensors</source>
            <volume>22</volume>
            <elocation-id>6299</elocation-id>
            <pub-id pub-id-type="doi">10.3390/s22166299</pub-id>
            <pub-id pub-id-type="pmid">36016060</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B13">
        <label>13.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Islam, M.R., Oliullah, K., Kabir, M.M., Alom, M. and Mridha, M.F. (2023) Machine Learning Enabled IoT System for Soil Nutrients Monitoring and Crop Recommendation. <italic>Journal of Agriculture and Food Research</italic>, 14, Article 100880. https://doi.org/10.1016/j.jafr.2023.100880 <pub-id pub-id-type="doi">10.1016/j.jafr.2023.100880</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.jafr.2023.100880">https://doi.org/10.1016/j.jafr.2023.100880</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Islam, M.R.</string-name>
              <string-name>Oliullah, K.</string-name>
              <string-name>Kabir, M.M.</string-name>
              <string-name>Alom, M.</string-name>
              <string-name>Mridha, M.F.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Machine Learning Enabled IoT System for Soil Nutrients Monitoring and Crop Recommendation</article-title>
            <source>Journal of Agriculture and Food Research</source>
            <volume>14</volume>
            <elocation-id>100880</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.jafr.2023.100880</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B14">
        <label>14.</label>
        <citation-alternatives>
          <mixed-citation publication-type="web">Ingle, A. (2020) Crop Recommendation Dataset. Kaggle. https://www.kaggle.com/datasets/atharvaingle/crop-recommendation-dataset</mixed-citation>
          <element-citation publication-type="web">
            <person-group person-group-type="author">
              <string-name>Ingle, A.</string-name>
            </person-group>
            <year>2020</year>
            <article-title>Crop Recommendation Dataset</article-title>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B15">
        <label>15.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Paithane, P.M. (2023) Random Forest Algorithm Use for Crop Recommendation. <italic>Journal of Engineering and Technology for Industrial Applications</italic>, 9, 34-41. https://doi.org/10.5935/jetia.v9i43.906 <pub-id pub-id-type="doi">10.5935/jetia.v9i43.906</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.5935/jetia.v9i43.906">https://doi.org/10.5935/jetia.v9i43.906</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Paithane, P.M.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Random Forest Algorithm Use for Crop Recommendation</article-title>
            <source>Journal of Engineering and Technology for Industrial Applications</source>
            <volume>9</volume>
            <pub-id pub-id-type="doi">10.5935/jetia.v9i43.906</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B16">
        <label>16.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Mahesh, P. and Soundrapandiyan, R. (2024) Yield Prediction for Crops by Gradient-Based Algorithms. <italic>PLOS ONE</italic>, 19, e0291928. https://doi.org/10.1371/journal.pone.0291928 <pub-id pub-id-type="doi">10.1371/journal.pone.0291928</pub-id><pub-id pub-id-type="pmid">39186769</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1371/journal.pone.0291928">https://doi.org/10.1371/journal.pone.0291928</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Mahesh, P.</string-name>
              <string-name>Soundrapandiyan, R.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Yield Prediction for Crops by Gradient-Based Algorithms</article-title>
            <source>PLOS ONE</source>
            <volume>19</volume>
            <pub-id pub-id-type="doi">10.1371/journal.pone.0291928</pub-id>
            <pub-id pub-id-type="pmid">39186769</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B17">
        <label>17.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Ramzan, S., Ali, B., Raza, A., Hussain, I., Fitriyani, N.L., Gu, Y., <italic>et al</italic>. (2024) An Innovative Artificial Neural Network Model for Smart Crop Prediction Using Sensory Network Based Soil Data. <italic>PeerJ Computer Science</italic>, 10, e2478. https://doi.org/10.7717/peerj-cs.2478 <pub-id pub-id-type="doi">10.7717/peerj-cs.2478</pub-id><pub-id pub-id-type="pmid">39650399</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.7717/peerj-cs.2478">https://doi.org/10.7717/peerj-cs.2478</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Ramzan, S.</string-name>
              <string-name>Ali, B.</string-name>
              <string-name>Raza, A.</string-name>
              <string-name>Hussain, I.</string-name>
              <string-name>Fitriyani, N.L.</string-name>
              <string-name>Gu, Y.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>An Innovative Artificial Neural Network Model for Smart Crop Prediction Using Sensory Network Based Soil Data</article-title>
            <source>PeerJ Computer Science</source>
            <volume>10</volume>
            <pub-id pub-id-type="doi">10.7717/peerj-cs.2478</pub-id>
            <pub-id pub-id-type="pmid">39650399</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B18">
        <label>18.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Ahmed, R.A., El-Shafai, W., Ahmed, Z.A., El-Rabaie, E.M. and El-Samie, F.E.A. (2025) High-Precision Crop Recommendation System with Stacking Ensemble Classifiers for Optimizing Agricultural Productivity. <italic>Scientific Reports</italic>, 15, Article No. 43377. https://doi.org/10.1038/s41598-025-09640-5 <pub-id pub-id-type="doi">10.1038/s41598-025-09640-5</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1038/s41598-025-09640-5">https://doi.org/10.1038/s41598-025-09640-5</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Ahmed, R.A.</string-name>
              <string-name>El-Shafai, W.</string-name>
              <string-name>Ahmed, Z.A.</string-name>
              <string-name>El-Rabaie, E.M.</string-name>
              <string-name>El-Samie, F.E.A.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>High-Precision Crop Recommendation System with Stacking Ensemble Classifiers for Optimizing Agricultural Productivity</article-title>
            <source>Scientific Reports</source>
            <volume>15</volume>
            <elocation-id>No</elocation-id>
            <pub-id pub-id-type="doi">10.1038/s41598-025-09640-5</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B19">
        <label>19.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Hasan, M., Marjan, M.A., Uddin, M.P., Afjal, M.I., Kardy, S., Ma, S., <italic>et al</italic>. (2023) Ensemble Machine Learning-Based Recommendation System for Effective Prediction of Suitable Agricultural Crop Cultivation. <italic>Frontiers in Plant Science</italic>, 14, Article 1234555. https://doi.org/10.3389/fpls.2023.1234555 <pub-id pub-id-type="doi">10.3389/fpls.2023.1234555</pub-id><pub-id pub-id-type="pmid">37636091</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3389/fpls.2023.1234555">https://doi.org/10.3389/fpls.2023.1234555</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Hasan, M.</string-name>
              <string-name>Marjan, M.A.</string-name>
              <string-name>Uddin, M.P.</string-name>
              <string-name>Afjal, M.I.</string-name>
              <string-name>Kardy, S.</string-name>
              <string-name>Ma, S.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Ensemble Machine Learning-Based Recommendation System for Effective Prediction of Suitable Agricultural Crop Cultivation</article-title>
            <source>Frontiers in Plant Science</source>
            <volume>14</volume>
            <elocation-id>1234555</elocation-id>
            <pub-id pub-id-type="doi">10.3389/fpls.2023.1234555</pub-id>
            <pub-id pub-id-type="pmid">37636091</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B20">
        <label>20.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Senapaty, M.K., Ray, A. and Padhy, N. (2024) A Decision Support System for Crop Recommendation Using Machine Learning Classification Algorithms. <italic>Agriculture</italic>, 14, Article 1256. https://doi.org/10.3390/agriculture14081256 <pub-id pub-id-type="doi">10.3390/agriculture14081256</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3390/agriculture14081256">https://doi.org/10.3390/agriculture14081256</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Senapaty, M.K.</string-name>
              <string-name>Ray, A.</string-name>
              <string-name>Padhy, N.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>A Decision Support System for Crop Recommendation Using Machine Learning Classification Algorithms</article-title>
            <source>Agriculture</source>
            <volume>14</volume>
            <elocation-id>1256</elocation-id>
            <pub-id pub-id-type="doi">10.3390/agriculture14081256</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B21">
        <label>21.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Janani, K., Alageswaran, R. and Amirtharajan, R. (2026) Dynamic SG-SKRDX Hybrid Framework for Precision Weather Forecasting and Crop Suitability in the Cauvery Delta. <italic>Scientific Reports</italic>, 16, Article No. 2042.</mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Janani, K.</string-name>
              <string-name>Alageswaran, R.</string-name>
              <string-name>Amirtharajan, R.</string-name>
            </person-group>
            <year>2026</year>
            <article-title>Dynamic SG-SKRDX Hybrid Framework for Precision Weather Forecasting and Crop Suitability in the Cauvery Delta</article-title>
            <source>Scientific Reports</source>
            <volume>16</volume>
            <elocation-id>No</elocation-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B22">
        <label>22.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Shastri, S., Kumar, S., Mansotra, V. and Salgotra, R. (2025) Advancing Crop Recommendation System with Supervised Machine Learning and Explainable Artificial Intelligence. <italic>Scientific Reports</italic>, 15, Article No. 25498. https://doi.org/10.1038/s41598-025-07003-8 <pub-id pub-id-type="doi">10.1038/s41598-025-07003-8</pub-id><pub-id pub-id-type="pmid">40664775</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1038/s41598-025-07003-8">https://doi.org/10.1038/s41598-025-07003-8</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Shastri, S.</string-name>
              <string-name>Kumar, S.</string-name>
              <string-name>Mansotra, V.</string-name>
              <string-name>Salgotra, R.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>Advancing Crop Recommendation System with Supervised Machine Learning and Explainable Artificial Intelligence</article-title>
            <source>Scientific Reports</source>
            <volume>15</volume>
            <elocation-id>No</elocation-id>
            <pub-id pub-id-type="doi">10.1038/s41598-025-07003-8</pub-id>
            <pub-id pub-id-type="pmid">40664775</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B23">
        <label>23.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Bakr, M.A., Khan, A.J., Khan, S.D., Zafar, M.H., Ullah, M. and Ullah, H. (2025) Evaluation of Learning-Based Models for Crop Recommendation in Smart Agriculture. <italic>Information</italic>, 16, Article 632. https://doi.org/10.3390/info16080632 <pub-id pub-id-type="doi">10.3390/info16080632</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3390/info16080632">https://doi.org/10.3390/info16080632</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Bakr, M.A.</string-name>
              <string-name>Khan, A.J.</string-name>
              <string-name>Khan, S.D.</string-name>
              <string-name>Zafar, M.H.</string-name>
              <string-name>Ullah, M.</string-name>
              <string-name>Ullah, H.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>Evaluation of Learning-Based Models for Crop Recommendation in Smart Agriculture</article-title>
            <source>Information</source>
            <volume>16</volume>
            <elocation-id>632</elocation-id>
            <pub-id pub-id-type="doi">10.3390/info16080632</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B24">
        <label>24.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Patel, K. and Patel, H.B. (2023) Multi-Criteria Agriculture Recommendation System Using Machine Learning for Crop and Fertilizers Prediction. <italic>Current</italic><italic>Agriculture</italic><italic>Research</italic><italic>Journal</italic>, 11, 137-149. https://doi.org/10.12944/carj.11.1.12 <pub-id pub-id-type="doi">10.12944/carj.11.1.12</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.12944/carj.11.1.12">https://doi.org/10.12944/carj.11.1.12</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Patel, K.</string-name>
              <string-name>Patel, H.B.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Multi-Criteria Agriculture Recommendation System Using Machine Learning for Crop and Fertilizers Prediction</article-title>
            <source>Current Agriculture Research Journal</source>
            <volume>11</volume>
            <pub-id pub-id-type="doi">10.12944/carj.11.1.12</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B25">
        <label>25.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Patel, K. and Patel, H.B. (2020) A State-of-the-Art Survey on Recommendation System and Prospective Extensions. <italic>Computers</italic><italic>and</italic><italic>Electronics</italic><italic>in</italic><italic>Agriculture</italic>, 178, Article 105779. https://doi.org/10.1016/j.compag.2020.105779 <pub-id pub-id-type="doi">10.1016/j.compag.2020.105779</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.compag.2020.105779">https://doi.org/10.1016/j.compag.2020.105779</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Patel, K.</string-name>
              <string-name>Patel, H.B.</string-name>
            </person-group>
            <year>2020</year>
            <article-title>A State-of-the-Art Survey on Recommendation System and Prospective Extensions</article-title>
            <source>Computers and Electronics in Agriculture</source>
            <volume>178</volume>
            <elocation-id>105779</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.compag.2020.105779</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B26">
        <label>26.</label>
        <citation-alternatives>
          <mixed-citation publication-type="web">Garg, S. (2025) Crop Dataset. Kaggle. https://www.kaggle.com/datasets/sashankgarg23/crop-dataset</mixed-citation>
          <element-citation publication-type="web">
            <person-group person-group-type="author">
              <string-name>Garg, S.</string-name>
            </person-group>
            <year>2025</year>
            <article-title>Crop Dataset</article-title>
          </element-citation>
        </citation-alternatives>
      </ref>
    </ref-list>
  </back>
</article>