Machine Learning Price Predictor: Ridge AR

The Machine Learning Price Predictor: Ridge AR indicator uses Regularized AutoRegression (Ridge AR) to provide stable, short-term price forecasting while minimizing overfitting in volatile market conditions. This tool is designed to help users visualize trend direction and momentum shifts through predictive modeling and color-coded segments.

Usage

The Usage section describes how the script can be used for market analysis. The indicator plots an in-sample "Fit Line" to show historical model accuracy and an out-of-sample "Forecast Line" to project future price paths.

The interpretation of the model is primarily driven by the trend segments and Bull/Bear labels. When upward momentum is confirmed based on the model's momentum weight calculation, a "Bull" label appears below the bar. Conversely, a "Bear" label appears above the bar when downward momentum is dominant. These signals are generated during recalculation cycles or continuously, depending on the selected Update Mode.

Details

The script implements a Ridge Regression model to solve for stable autoregressive coefficients. By adding a regularization parameter ($\lambda$), the model handles correlated datasets and noisy price action more effectively than standard linear regression.

The workflow consists of several algorithmic stages:

1. Prefiltering: The raw price data is smoothed using EMA or the Ehlers SuperSmoother to reduce high-frequency noise before the model is trained.
2. Training: The model uses a defined window of historical bars to calculate the relationship between past lags (AR Order) and current prices.
3. Regularization: The Ridge Strength coefficient is applied to the normal equations to prevent the model from becoming overly sensitive to small price fluctuations.
4. Projection: The forecast is generated by integrating the trained coefficients with a damping factor, recent momentum, and mean reversion tendencies.

Settings

Ridge AR Settings

• Training Window: The number of historical bars used to train the model.
• Forecast Horizon: The number of bars into the future the forecast line will project.
• AR Order: The number of lags used as features for the autoregressive model (max 5).
• Ridge Strength (λ): The regularization coefficient; higher values increase model stability but may reduce sensitivity.
• Damping Factor: Controls the exponential decay rate of the trend in the forecast.
• Trend Length: The period used to estimate current volatility and trend direction.
• Momentum Weight: Adjusts how strongly the most recent price move influences the forecast.
• Mean Reversion: Determines the intensity of the projected price's pull back toward the training mean.

Data Processing

• Prefilter: Selection between None (raw price), EMA, or SuperSmoother for data smoothing.
• EMA Length / SuperSmoother Length: The lookback period for the selected smoothing filter.

Display Settings

• Update Mode: "Lock" builds the model once; "Update Once Reached" rebuilds after the forecast horizon expires; "Continuous" updates on every bar.
• Forecast Color: Sets the color for the out-of-sample projection line.
• Bullish/Bearish Colors: Sets the colors for the trend segments and labels.

FAQ

How do I interpret the Forecast Line? The Forecast Line represents a statistical projection based on recent price history and autoregressive coefficients. It is intended to show potential directional bias rather than a guaranteed price target.

What is the benefit of Ridge over standard Linear Regression? Ridge Regression includes a penalty for large coefficients (regularization), which helps prevent the model from overfitting to noise, leading to more consistent projections in volatile markets.

How do I access this indicator? You can get access on the LuxAlgo Library for charting platforms like TradingView, MetaTrader (MT4/MT5), and NinjaTrader for free.