Generate predictive trading signals from OHLCV data with walk-forward validation.
npx skills add https://github.com/hxhyyy/Vibe-Trading --skill ml-strategy-hxhyyy
Please help me install this Agent Skill. Skill: ml-strategy Source: https://github.com/hxhyyy/Vibe-Trading/tree/main/agent/src/skills/ml-strategy Command: npx skills add https://github.com/hxhyyy/Vibe-Trading --skill ml-strategy-hxhyyy
This skill addresses the complexity of building reliable, leak-proof machine learning models for financial time-series data by providing a standardized, safety-first pipeline.
Use the ml-strategy skill to generate trading signals for the provided OHLCV data using the random forest model.
High-intent search queries and answers about installing and using this skill.
You can generate predictive trading signals from OHLCV data by training machine learning models like RandomForest or GradientBoosting on engineered technical indicators. This pipeline performs feature engineering and model training to output actionable predictive signals for quantitative trading research.
Walk-forward validation prevents look-ahead bias by training and testing models on expanding or sliding windows sequentially. This ensures the model only evaluates on unseen future data chronologically, maintaining data integrity for reliable financial time-series forecasting.
Yes, you need scikit-learn, pandas, and numpy to build quantitative trading strategies with this approach. These dependencies are required to perform feature engineering, train machine learning models, and manipulate OHLCV financial data arrays for signal generation.
The best way to prevent data leakage is using a safety-first pipeline with walk-forward validation. By automatically applying expanding or sliding windows to separate training and testing data chronologically, it ensures future data does not influence past predictions.
Yes, you can use Ridge regression alongside technical indicators like RSI and Bollinger Bands. The pipeline includes pre-built indicators with division-by-zero protection and supports selecting between RandomForest, GradientBoosting, or Ridge models for trading signal generation.
Machine learning trading strategies often fail on unseen data due to look-ahead bias from improper validation. Implementing walk-forward validation and using built-in division-by-zero protection during feature engineering ensures robust model generalization and prevents future data leakage.
omicverseClean, reshape, normalize and engineer features from CSV/TSV/Parquet inputs using pandas and numpy locally.
cyoteeCompute TWAP data from Uniswap V3 on-chain oracle observations.
shushuznCompute SHAP values and generate interpretability visualizations for machine learning models.
kishorkukrejaForecast demand using LSTM, GRU, and Transformer deep learning models.
