Multi-Indicator Technical & ML Sector Forecasting FastAPI

Upload or select sector price data, train Random Forest or XGBoost classifiers on 12 technical indicators, backtest a consecutive-signal entry strategy, and generate a downloadable PDF + CSV report per sector. Optionally enable regression mode for continuous return forecasts.

📄 Published Reports

Loading reports…

⚙ Run Analysis

📂 Use Existing File
Files already on the server
⬆ Upload New File
Upload your own .xlsx file
🌲 Random Forest
300 estimators · no extra deps · robust
⚡ XGBoost
Gradient boosting · often higher accuracy

📋 Past Reports

Loading…

📖 Methodology Reference

Hover each chip for details. All computed from the Close price column.

RSI MACD Coppock Curve Bollinger Bands KST PPO Stochastic %K/%D Williams %R MAS CLV MA50 / MA200 ROC
Classification Target Binary: 1 if 5-day forward return > 0, else 0. 80/20 chronological train/test split — no look-ahead bias.

Feature Set All 12 indicator values + lagged versions (RSI_lag1, MACD_lag1, Coppock_lag1) + binary threshold signals (RSI over 30/50/70, MACD crossover, Coppock turn-up, Coppock zero-cross).
Models Random Forest (300 estimators, n_jobs=-1) or XGBoost (300 rounds, binary:logistic). Feature importance via permutation importance on the held-out test set.

Regression Mode Same features, continuous 5-day return as target. Reports R², MAE, RMSE + a single-point forecast for the next period.
Entry Signal 3 consecutive days where the classifier predicts a positive 5-day return. Entered at the next day's close price.

Exit Fixed 20 trading day hold (~1 calendar month). No stop-loss, no take-profit, no transaction costs.
Metrics
  • ML Accuracy — classification accuracy on test set
  • Cumulative Return — geometric product of all trade returns
  • Win Rate — % of trades with positive return
  • Num Trades — total entry signals triggered
  • Forecast Prob Up — model confidence on latest data point