Machine Learning Random Forest Strategy

The Machine Learning Random Forest Strategy indicator utilizes ensemble learning techniques to forecast market direction and price magnitude. By combining multiple decision stumps into a shallow Random Forest model, the tool provides a probabilistic outlook on whether the next move is likely to be bullish or bearish, alongside an estimated expected return.

Usage

The indicator can be used to identify high-probability reversal or continuation setups based on machine learning classifications.

• Bullish Signals: Triggered when the bullish probability exceeds the defined threshold (default 60%) and the cooldown period has passed.
• Bearish Signals: Triggered when the bearish probability (1 - Bull Prob) exceeds the threshold.
• Confirmation: Stronger signals occur when the Classifier (direction) and the Regression Forest (magnitude/expected return) align. For example, a high bullish probability coupled with a positive expected return indicates higher conviction.
• Adaptive Threshold: If enabled, the tool automatically raises the required probability threshold during periods of poor performance, filtering out lower-quality signals in difficult market regimes.

Details

This script implements a depth-1 Random Forest using "decision stumps" to stay within the execution limits of Pine Script while maintaining ensemble logic.

• Bootstrap Sampling: Each tree is trained on a random subset of historical data.
• Information Gain: The model performs a threshold search using Gini Impurity to select the most effective splits for its predictions.
• Rolling Window: The model continuously retrains on the most recent N bars (default 60), allowing it to adapt to non-stationary market conditions without lookahead bias.
• Features: The model evaluates three primary inputs: an Anchor Oscillator (RSI, MFI, Stoch, or Z-Score), Trend Correlation (price vs. time), and Normalized Momentum.

Settings

Model & Logic

• Anchor Oscillator/Length: Sets the primary momentum or mean reversion feature and its lookback period.
• Training Window (bars): Determines how much recent history the model uses for the rolling retraining process.
• Number of Trees: Controls the number of decision stumps in the ensemble.
• Signal Probability Threshold: The baseline probability required to trigger a signal.
• Adaptive Threshold: Toggles the self-correcting logic that tightens signal requirements based on recent performance.

Filters & Visuals

• Signal Cooldown: The minimum number of bars required between signals of the same direction to prevent clustering.
• Enable ADX Trend Filter: Filters signals based on a minimum ADX level to prioritize trending environments.
• TP/SL Settings: Adjusts ATR-based target price and stop loss levels for signal visualization.
• Show Backtest Table: Displays a summary of win rate, profit factor, and cumulative PnL on the chart.

FAQ

How do I access the Machine Learning Random Forest Strategy?

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

What features does the model use to predict price?

The model evaluates four normalized features: the selected Anchor Oscillator, Trend Correlation, Momentum normalized by ATR, and recent volatility.

Why does the probability threshold change on my chart?

If "Adaptive Threshold" is enabled, the indicator increases the required probability for a signal if recent trades have been unsuccessful, signaled by an orange marker in the statistics table.