FinSight: Multitasks AI-Based Platform for Financial Market Prediction and Analysis

Abstract

This thesis presents FinSight AI, a multi-task financial decision support platform that applies machine learning and deep learning techniques to stock price prediction, trend classification, sales forecasting, credit risk assessment, and financial sentiment analysis. Several models were trained and evaluated on real-world datasets. The results show that deep learning does not always outperform classical machine learning and the used metrics proved the quality of the model. GRU achieved the best stock price prediction, Linear Regression performed best for trend classification, the Neural Ensemble led credit risk assessment, and fine-tuned FinBERT achieved 83.08% accuracy in sentiment analysis. All models were integrated into an interactive Streamlit web application.