Trading Bot
Dashboard

Trading Bot Dashboard

Autonomous trading powered by time series forecasting models and sentiment analysis of financial news.

About This Project

This trading bot is a university research project for the DSAI course. It uses time series forecasting models and sentiment analysis of financial news to make autonomous trading decisions on the Alpaca paper trading platform.

Executive Interpretation
Key findings derived from the experimental report.

Outperformance vs Buy-and-Hold

The RL agent consistently outperforms buy-and-hold, especially on high volatility symbols like MSTR and BA.

Reward Metric Matters

Sortino reward is most stable for defensive names (MSFT, KO, GLD), while Profit excels on high-beta names like NVDA and MSTR.

Sentiment Stabilizes

Sentiment features smooth returns for lower volatility assets but can dampen performance on momentum-heavy stocks.

Narrative Summary

The hybrid Chronos + FinBERT + RL system shows measurable alpha in volatile regimes. It excels at controlling drawdowns while maintaining exposure during stable uptrends. Sentiment improves stability for conservative assets, while pure price signals dominate in speculative names (e.g., NVDA, MSTR).

Caveats & Checks

  • Extreme returns (e.g., MSTR) should be revalidated for data leakage or test overlap.
  • Sentiment lag can hurt performance in hype-driven rallies.
  • Reward selection changes the risk profile more than the forecast model choice.
Live Performance
Monitor your portfolio in real-time. Track positions, P&L, and recent trades.
Statistical Results
Compare model performance, view backtesting results, and understand why our model works.
Trading Bot
Navigation