TITLE:
TriFusion Stacking (TFS) for Stock Prediction: Integrating Sentiment, Technical, and Fundamental Analysis
AUTHORS:
Shawn Tang
KEYWORDS:
Stock Price Prediction, Fusion Model, Stacking, Ensemble Methods
JOURNAL NAME:
Journal of Computer and Communications,
Vol.13 No.12,
December
26,
2025
ABSTRACT: This paper proposes a hybrid AI framework that integrates technical indicators, fundamental data, and financial news sentiment into a stacked ensemble learning model. The ensemble combines 14 algorithms, including decision tree, gradient boosting machine (GBM), gated recurrent unit (GRU), k-nearest neighbor (KNN), lasso regression, linear regression, long short-term memory (LSTM), multi-layer perceptron (MLP), random forest, ridge regression, recurrent neural network (RNN), spiking neural network (SNN), support vector machine (SVM), and temporal convolutional network (TCN). Each prediction is fused through an MLP to generate final outputs. This stacking approach yielded a maximum return of $4061 from an initial investment of $1000. Our results show that simple linear models generally outperform deep learning models in stock predictions, and single stocks are more predictable than other volatile assets or index stocks.