Machine Learning Moving Average

The Machine Learning Moving Average indicator provides an adaptive trend-following and forecasting framework by categorizing market behavior into statistical clusters. It utilizes a pseudo-machine learning approach to identify bullish, neutral, and bearish regimes, calculating a dynamic moving average and price predictions based on historical feature proximity.

Usage

The Usage section describes how the script can be used to identify market regimes and forecast potential price movements. Traders can observe the color-coded candles and background zones to determine the current market state: green for bullish, red for bearish, and blue/gray for neutral.

The indicator provides a forward-looking projection (predicted price) based on the selected ML method. The "Confidence %" helps filter these signals; a high confidence rating suggests the current market behavior strongly aligns with a specific historical cluster, whereas low confidence indicates a potential regime transition or market noise.

• Trend Following: Use the ML Moving Average (MLMA) line as a dynamic support/resistance level.
• Regime Filtering: Adjust trading strategies based on the current cluster (e.g., trend-following in Bull/Bear clusters and mean-reversion in Neutral clusters).
• Forecasts: Use the predicted price and horizon to set realistic profit targets or time exits.

Details

The script operates through a multi-stage process involving feature engineering, percentile clustering, and predictive modeling:

• Feature Engineering: It generates a "Composite Feature" by calculating Z-scores for RSI, CCI, ATR, Volume, and Momentum. This normalizes different data types into a single statistical metric.
• Percentile Clustering: Unlike traditional K-Means, this script uses percentile thresholds (Lower and Upper) to define the boundaries of market regimes. It sorts historical data to find the 10th and 90th percentiles (adjustable), effectively "training" the model on the extremes of past volatility and momentum.
• Pseudo-ML Prediction: It calculates the expected return by analyzing the historical average returns of the current cluster.
• ML Moving Average: The adaptive MLMA adjusts its smoothing factor based on the current regime and prediction horizon, reacting faster during high-confidence trend shifts and smoothing out during neutral periods.

Settings

Feature Engineering

• RSI/CCI/ATR/Volume/Momentum Lengths: Sets the lookback period for each individual technical indicator used in the composite feature.
• Use [Feature] Toggles: Allows users to include or exclude specific indicators from the final clustering calculation.

Clustering Settings

• Number of Clusters: Choose between 2 (Bullish/Bearish) or 3 (Bullish/Neutral/Bearish) market states.
• Lookback Period: The window of historical data used to determine percentile thresholds.
• Lower/Upper Percentile: Defines the statistical boundaries for the Bearish and Bullish clusters.

ML Prediction

• Prediction Horizon: The number of bars into the future the prediction aims to forecast.
• Training Window: The number of recent data points used to calculate average cluster returns.
• ML Method:
  • Weighted Average: Predicts based on the simple average return of the current cluster.
  • Cluster Momentum: Adjusts the prediction based on the rate of change within the cluster.
  • Regime Transition: Interpolates between the primary and secondary nearest clusters to predict transitions.

Visual Settings

• Show Prediction Numbers/Confidence: Toggles the visibility of the price target and confidence percentage.
• Show Cluster Zones/MLMA: Enables or disables the background regime coloring and the adaptive moving average line.
• Paint candles by trend: Colors the price bars according to the active cluster.

FAQ

How should I interpret the Confidence %? The confidence percentage represents how closely the current market data fits into its assigned cluster compared to the next closest one. Higher values suggest a stable regime, while lower values often precede a trend change or a shift into a neutral state.

Can I use this for scalping or long-term investing? Yes, the indicator is timeframe-agnostic. By adjusting the "Lookback Period" and "Prediction Horizon," users can optimize the tool for high-frequency scalping or daily/weekly trend analysis.

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