How bhumikaaggwl/equity-return-prediction-alpha-signal-model works
This is an equity-return research and alpha-signal platform: it ingests market, fundamentals, and news data, builds features and excess-return targets, trains and evaluates models, then produces signals and research views. The main batch workflow is explicit in main.py; the README documents SHAP explainability and signal-basket risk analysis. Some feature and dashboard implementations were not excerpted, so their documented capabilities are shown without assuming unverified internal calls.
Main entry points: storage.py SQLite research store · trainer.py Model training and comparison · main.py Pipeline runner · builder.py Feature and target builder
Components 25
Data ingestion
- Price ingestion
data/fetcher.py - Fundamentals ingestion
data/fetcher.py - News ingestion
data/news_fetcher.py - Market data provider
- SQLite research store
data/storage.py
Feature engineering
- Feature and target builder
features/builder.py - Technical indicators
features/technical.py - Fundamental ratios
features/fundamental.py - Valuation metrics
features/valuation.py - FinBERT sentiment
features/sentiment.py - FinBERT model
Model research
- Model training and comparison
models/trainer.py - Scratch decision tree
models/decision_tree.py - Scratch XGBoost
models/xgboost_model.py - Prediction metrics
evaluation/metrics.py - Walk-forward validation
backtest/walkforward.py
Signals and research
- BUY HOLD SELL signals
recommendation/signal.py - SHAP explainability
explainability/shap_analysis.py - Signal basket risk
evaluation/risk.py - Interactive dashboard
dashboard/app.py - Research report export
export_report.py
Other components
- Research user
- Pipeline runner
main.py - Universe and settings
config.py - Logging
src/utils/logger.py
Connections 41
- Research user → Pipeline runner: starts pipeline
README.md - Research user → Interactive dashboard: explores research
README.md - Pipeline runner → SQLite research store: initializes
main.py - Pipeline runner → Price ingestion: fetches prices
main.py - Pipeline runner → Fundamentals ingestion: fetches fundamentals
main.py - Pipeline runner → News ingestion: fetches headlines
main.py - Pipeline runner → Model training and comparison: trains models
main.py - Pipeline runner → Walk-forward validation: runs validation
main.py - Price ingestion → Market data provider: downloads prices
README.md - Fundamentals ingestion → Market data provider: fetches fundamentals
README.md - News ingestion → Market data provider: fetches news
README.md - Price ingestion → SQLite research store: writes prices
data/fetcher.py - Fundamentals ingestion → SQLite research store: writes records
data/fetcher.py - News ingestion → SQLite research store: writes headlines
data/news_fetcher.py - Feature and target builder → SQLite research store: reads prices
features/builder.py - Feature and target builder → Technical indicators: computes indicators
features/builder.py - Feature and target builder → Fundamental ratios: computes ratios
features/builder.py - Feature and target builder → Valuation metrics: computes valuation
features/builder.py - Feature and target builder → FinBERT sentiment: merges sentiment
features/builder.py - FinBERT sentiment → SQLite research store: reads headlines
features/sentiment.py - FinBERT sentiment → FinBERT model: scores headlines
features/sentiment.py - Model training and comparison → Feature and target builder: builds training data
models/trainer.py - Model training and comparison → Scratch decision tree: trains scratch tree
models/trainer.py - Model training and comparison → Scratch XGBoost: trains scratch model
models/trainer.py - Model training and comparison → Prediction metrics: scores predictions
models/trainer.py - Model training and comparison → SQLite research store: stores metrics
models/trainer.py - Walk-forward validation → Feature and target builder: builds test data
backtest/walkforward.py - Walk-forward validation → Prediction metrics: evaluates folds
backtest/walkforward.py - Walk-forward validation → BUY HOLD SELL signals: counts fold signals
backtest/walkforward.py - Model training and comparison → BUY HOLD SELL signals: generates saved signals
models/trainer.py - Interactive dashboard → SQLite research store: reads research data
dashboard/app.py - Interactive dashboard → Feature and target builder: builds features
dashboard/app.py - Interactive dashboard → BUY HOLD SELL signals: generates signals
dashboard/app.py - Interactive dashboard → Prediction metrics: computes metrics
dashboard/app.py - Interactive dashboard → Walk-forward validation: runs backtest
dashboard/app.py - Interactive dashboard → SHAP explainability: explains predictions
README.md - Research report export → Signal basket risk: reports risk
export_report.py - Signal basket risk → SQLite research store: reads signal data
evaluation/risk.py - Pipeline runner → Logging: logs pipeline
main.py - Model training and comparison → Logging: logs training
models/trainer.py - Universe and settings → SQLite research store: sets database path
data/storage.py