TITLE:
Forecasting Stock Prices for Automated Trading Using Machine Learning with Historical Data
AUTHORS:
Matthew Shen, Odysseas Drosis
KEYWORDS:
Stock Price Forecasting, Automated Trading, Machine Learning, Historical Data, Financial Prediction, Algorithmic Trading, Time Series Analysis
JOURNAL NAME:
Open Journal of Applied Sciences,
Vol.15 No.9,
September
17,
2025
ABSTRACT: Stock markets are notoriously complex and volatile, which makes accurate price prediction a necessity for investor decision-making, risk mitigation, and profitability. This study develops and evaluates an AI-driven trading bot using historical stock data to forecast short-term price movements. The core research question examines whether machine learning can reliably predict prices to generate profitable automated trades. The hypothesis posits that effectively trained ML models can identify patterns to enable superior trading strategies compared to traditional methods. Four models—Linear Regression, Random Forest, Decision Tree, and MLP Regressor—were trained and assessed using Mean Squared Error and a custom Mean Absolute Percentage Error metric. The Random Forest model’s predictions directed a simulated trading bot executing buy/sell decisions over 1200 time points, starting with a user-specified amount of capital and incorporating technical indicators and risk management rules. The simulation showcased a tangible potential for profit by consistently achieving considerable returns with reduced risk. These findings strongly support the use of ML for automated trading and offer investors a powerful tool to optimize portfolio performance.