CAT — AI-Powered CSE Trading Suggestion System

A daily buy/sell/hold suggestion system for the Colombo Stock Exchange, fusing an XGBoost model trained on candlestick data across roughly 100 stocks with FinBERT news-sentiment scoring and LLM-extracted fundamentals from annual and quarterly reports into a single reasoned recommendation.

I designed a rigorous evaluation pipeline to keep the results honest: time-aware train/test splits to avoid look-ahead bias, per-year backtesting, closed-loop suggestion tracking through to predicted exit, and weekly model retraining.