Smart Stock Market Analysis Using Multi-Source Learning Models

Abstract

This project presents a web-based application designed to predict stock market trends by combining financial news with historical stock data. The system processes large datasets and extracts useful features such as sentiment scores and event-based vectors. Different machine learning models, including Support Vector Machine (SVM), a proposed multi-instance learning approach, and an extended XGBoost algorithm, are applied and compared using evaluation metrics like accuracy, precision, recall, and F-score. The results show that integrating textual sentiment with numerical stock data improves prediction performance. Among all models, the extended XGBoost algorithm achieves the best results. The application provides a simple interface that allows users to load datasets, perform feature extraction, train models, and generate predictions, making it useful for investors who want data-driven insights.