ml-strategy

What problem does it solve?

ML strategies on trading data often suffer from data leakage, inf-laden features, and unbounded predictions, so this Skill validates OHLCV inputs, sanitizes every feature, and keeps outputs clipped while relying solely on historical windows for training.

Core Features & Use Cases

• Data hygiene and feature engineering: OHLCV columns are validated, momentum/volatility/RSI ratios are guarded against division-by-zero, and inf values are scrubbed before modeling.
• Walk-forward training with model choice: Expanding or sliding training windows retrain RandomForest, GradientBoosting, or Ridge classifiers every configurable interval, and the prediction is normalized to a [-1.0, 1.0] confidence range.
• Use Case: Feed the Skill your universe of daily OHLCV time series, let it generate probability-based signals for each symbol, and consume those signals in your systematic rebalancing engine.

Quick Start

Run the ml-strategy on recent OHLCV data to receive normalized buy and sell signals with walk-forward ML.