It's time to put your temporal intelligence to the ultimate test. In this project, you will build a high-performance stock forecasting engine.
1The Hybrid Ensemble
Real-world financial data is complex. To capture every pattern, you will build a Hybrid Ensemble. You'll use XGBoost to process technical indicators (like RSI and MACD) and categorical features (day of week). Simultaneously, you'll use an LSTM to process the raw price sequence to capture long-term momentum. By combining their predictions, you create a model that is significantly more robust than any single architecture.
2Walk-Forward Integrity
A stock model that can't survive a backtest is a liability. You will implement a rigorous Walk-Forward Validation scheme across three years of historical data. You will ensure that at no point does your model 'see' future prices. You'll calculate not just error (RMSE), but also Directional Accuracy—how often your model correctly predicts if the price will go up or down, regardless of the magnitude.
3The Bottom Line
To graduate, you must demonstrate that your model is usable. You will build a simulated trading strategy based on your forecasts and calculate its Sharpe Ratio and Max Drawdown. This level of professional evaluation is what separates an AI researcher from a quantitative developer. You will prove that your temporal intelligence can provide consistent, risk-managed value in a volatile world.
