01 / PROBLEM STATEMENT
Why this work exists
Market experiments are difficult to audit when collection, cleaning, features, models, and backtests live in disconnected notebooks. This educational application makes each stage visible in one interface.
Data and source material
- Configurable NASDAQ tickers downloaded with yfinance
- OHLCV history from a configurable start date
- Technical, calendar, and lagged features
02 / SOLUTION OVERVIEW
From input to usable output
A modular application keeps each analysis stage visible and exportable.
- 01
Download and inspect raw market data
- 02
Handle missing values and optional z-score anomalies
- 03
Create moving averages, RSI, MACD, Bollinger Bands, ATR, volatility, calendar, and lag features
- 04
Compare Decision Tree, Random Forest, and XGBoost classifiers
- 05
Backtest buy-and-hold, moving-average crossover, and RSI mean reversion
03 / TECHNICAL ARCHITECTURE
Implemented components
yfinance → quality → features → model comparison → strategy backtests
04 / METHODOLOGY & IMPLEMENTATION
Decisions that shape the system
- PCA and t-SNE provide exploratory dimensionality views.
- The documented model path uses an 80/20 split and StandardScaler.
- Backtests report total return, equity path, maximum drawdown, and final value.
Implemented capabilities
- Data cleaning
- Technical indicators
- PCA and t-SNE
- Three classifiers
- Three backtests
05 / RESULTS & EVIDENCE
What the source supports
- Eight implemented modules cover collection through backtesting.
- The repository does not publish a stable tracked metric table, so no model or strategy performance figure is claimed.
Results preserve the repository’s units, evaluation framing, and stated limitations. Missing evidence is not converted into a claim.
06 / CHALLENGES & LIMITATIONS
Where the evidence stops
- The documented 80/20 split is educational and not a walk-forward production design.
- Backtests omit commissions and slippage.
- Data is historical only, and t-SNE is limited to roughly 5,000 observations.
07 / WHAT THIS DEMONSTRATES
Engineering signal
End-to-end analytical workflow design, technical feature engineering, model comparison, and explicit separation between a teaching lab and production research.
08 / SOURCE & RELATED WORK
Trace the evidence
58f8d5402a247a9b899d2c788cf3d9cb2e1bc794 ↗README.md · RAPPORT_SYNTHESE.md · GUIDE_DETAILLE.md · app.py