Danario Edgar Master’s Thesis Defense, Wednesday, April 29, 2026 @ 3:00 pm Central Time

COMMITTEE CHAIR: Dr. Md Hossain Shuvo

TITLE: INTELLITRADEAI: AN AI-POWERED MULTI-ASSET TRADING PLATFORM USING ENSEMBLE MACHINE LEARNING, CHART PATTERN RECOGNITION AND TRI-SIGNAL FUSION

ABSTRACT: IntelliTradeAI is an AI-powered algorithmic trading platform covering 259 assets across equities, ETFs, cryptocurrencies, and forex. The system fuses three independent signal sources — an ensemble of Random Forest and XGBoost classifiers trained on 70 volatility-aware technical features, a chart pattern recognition engine covering 20 classical formations, and a financial news intelligence module — through a convergence-gated architecture that generates signals only when multiple layers align. A market regime filter leveraging the CBOE VIX, the Fear and Greed Index, and per-symbol trend classification suppresses counter-regime signals, materially improving risk-adjusted performance. Walk-forward backtesting across 36 out-of-sample months (January 2022–December 2024) yields directional accuracy of 85.2% on equities, 96.3% on ETFs, 71.4% on forex, and 54.7% on cryptocurrencies. Over 847 backtested trades, the system achieves a Sharpe Ratio of 1.74, Calmar Ratio of 2.31, maximum drawdown of -15.1%, and annualized return of 34.9%, versus 15.2% for SPY buy-and-hold. Ablation testing confirms regime filtering is central — its removal alone drops the Sharpe Ratio from 1.74 to 0.83. Post-publication enhancements include replacing the lexicon-based sentiment module with a FinBERT transformer, improving sentiment accuracy from ~75% to over 90%. A deep learning benchmark revealed the Transformer architecture achieved 61.4% standalone accuracy versus the RF+XGBoost baseline of 55.8%; however, the full IntelliTrade system still reaches 85.2%, demonstrating that feature engineering and multi-signal fusion outperform any single architecture. Live paper trading via the Alpaca API (April 2–9, 2026) produced a 78.0% win rate across 381 trades and a +2.04% eight-day return. The platform is delivered as a Streamlit web application with real-time signal scanning, portfolio tracking, and SHAP-based explainability — with EMA alignment, ADX strength, and ATR-normalized volatility identified as top feature drivers. All code is open source and available at: https://github.com/djedgar1018/IntelliTradeAI.git

Keywords: Tri-Signal Fusion, Machine Learning, AI-Powered trading, Explainable AI

Room Location: S. R. Collins. Room 111: CS Main Conference Room.