what extent could Machine Learning Revolutionize the Approach to Stock Market and Trade Predictions?

This study evaluates the predictive performance of four machine learning models with different types of supervision. Logistic regression, Central Neural Network (CNN), random forest, standard Support Vector Machine (SVM), and sparse SVM on U.S. equity data spanning Q1 2024 to Q2 2025. Using 30 technical and sentiment-based indicators, each model was assessed using accuracy, precision, recall, F1-score, and AUC. Sparse SVM achieved the highest overall performance, with 87.4% accuracy, 85.1% precision, 83.9% recall, 84.5% F1-score, and 91.2% AUC, while selecting only 7 features. These results indicate that sparse SVM offers superior predictive power and interpretability, demonstrating that machine learning, particularly models with embedded feature selection, can substantially improve the precision and efficiency of stock market forecasting.