Utilizing Deep Learning and Machine Learning for Predicting Stock Market Trends with Multivariate and Persistent Data

Abstract

The future volatility of equity markets remains unpredictable, posing risks in trend forecasting. This study minimizes such risk using machine-learning and deep-learning models on Tehran Stock Exchange data from the IT and Banking sectors. Eleven ML algorithms—Decision Tree, Random Forest, AdaBoost, XGBoost, SVC, Naïve Bayes, KNN, Logistic Regression, and ANN—are com-pared with deep models RNN and LSTM. Using ten lagged indicators, models are trained under two strategies: continuous data and binary classification. Results show that deep-learning models out-perform classical ones, with LSTM achieving an F1-score of 0.91 and RNN 0.88, compared to XGBoost (0.80) and Random Forest (0.78). Deep models capture temporal dependencies effectively, improving trend prediction stability. Recent studies affirm these results, showing deep architec-tures’ superiority in financial time-series forecasting, making LSTM a strong candidate for real-time market analysis.