Machine Learning & AI Strategies Pack

Overview

This pack includes 26 AI & ML strategies for modern systematic traders. Each strategy ships with full Python code, a theory‑first walkthrough, a developer manual, and a backtest harness you can adapt instantly.

What’s Inside: 26 Unique Strategies

• Kalman-Noise-Adaptive Trend Filter

  Estimates price trends using a Kalman filter that adaptively tunes its noise covariance based on rolling volatility.

• Autoencoder Denoising Residuals

  A neural network autoencoder reconstructs price data; signals are derived from the reconstruction errors (residuals).

• Contrastive-Predictive Coding Reversion

  Uses CPC as a proxy autoencoder to build latent features and trade reversion from latent space distances.

• Elastic-Net on Lagged Indicators

  Applies Elastic Net regression to a set of lagged technical indicators to generate predictive signals.

• Gaussian Mixture Cluster-Reversion

  Fits Gaussian Mixture Models (GMM) to identify market states and trades mean-reversion from cluster extremes.

• Kalman Filter Fair-Value Reversion

  A Kalman filter estimates an asset's fair value, with trades triggered by deviations from the observed price.

• Kernel Regression Fair-Value

  Uses nonlinear kernel ridge regression for fair-value estimation and subsequent mean-reversion trades.

• ATR-Scaled Trailing-Stop Momentum

  A classic momentum strategy augmented with dynamic trailing stops scaled by the Average True Range (ATR).

• Donchian Channel Breakout with Dynamic Bands

  A breakout strategy using Donchian Channels with band widths that adapt based on ATR.

• Kalman-Filter Trend Estimation

  A classic Kalman filter setup to estimate a latent trend and slope, used for crossover-based signals.

• Keltner Channel Breakout with Dynamic Bands

  A breakout strategy using Keltner Channels with adaptive, ATR-based bands.

• Machine-Learning-Augmented Momentum

  Enhances standard momentum features with an XGBoost model to classify and filter entry signals.

• ML-Probability Threshold Momentum

  Trades momentum signals only when an ML model's classification confidence exceeds a predefined threshold.

• Principal-Component Momentum

  Uses PCA on a basket of assets to trade momentum on the primary, market-wide component.

• Regime-Adaptive Momentum with HMM

  A Hidden Markov Model detects the market regime and adjusts momentum parameters accordingly.

• Skewness- & Kurtosis-Adjusted Momentum

  Filters momentum signals by avoiding trades during periods of high skewness or kurtosis (fat tails).

• Volatility-Scaled Time-Series Momentum

  Standard time-series momentum signals that are scaled by exponentially weighted volatility.

• Adaptive Kalman Filter Strategy

  A Backtrader implementation of an adaptive Kalman filter system with dynamic state updates.

• Decision Tree EMA Crossover Strategy

  A Decision Tree classifier that predicts moves based on features derived from EMA crossovers.

• Dual-Regime Weekly Strategy

  Switches between a trending and a ranging logic model on a weekly basis.

• Ensemble Strategy with Fixed Weights

  Combines signals from multiple sub-strategies using predefined, fixed weights.

• HMM Trend Regime Strategy

  A Hidden Markov Model that infers the underlying trend regime to generate trading signals.

• Hurst-Filtered Trend Strategy

  Uses the rolling Hurst exponent to determine whether to engage a trend-following strategy.

• Isolation Forest Anomaly Strategy

  An anomaly detection model (Isolation Forest) that identifies and trades on abnormal market patterns.

• Moving Average Channel Strategy

  A simple channel breakout strategy based on moving averages with built-in logging.

• Empirical-Mode Decomposition Channels

  Decomposes a signal into Intrinsic Mode Functions (IMFs) to create a noise envelope for breakout trades.

What You Get in the Pack

• Full Python source code (.py) for all 26 strategies
• A complete developer manual for quick reference