KNN Machine Learning Momentum Indicator

The KNN Machine Learning Momentum Indicator tool utilizes a K-Nearest Neighbors algorithm combined with dimensionality reduction to estimate short-term market momentum and identify high-probability trend shifts. It provides traders with a non-linear approach to momentum analysis by comparing current market features against historical patterns.

Usage

The Usage section describes how the script can be used to identify market entries and assess trend conviction. The indicator outputs visual signals and dynamic bar coloring to represent the model's confidence levels.

• Major Signals: Large shapes accompanied by labels (Bull/Bear) indicate signals that align with the selected EMA trend filter. These represent high-conviction momentum shifts.
• Minor/Counter Signals: Smaller shapes without labels indicate potential signals that run counter to the primary trend. These are often used for mean-reversion strategies or cautious entries.
• Dynamic Bar Coloring:
  • Bright Cyan/Red: High-confidence predictions where the KNN engine detects a strong similarity to historical profitable momentum shifts.
  • Slate Gray/Neutral: Low-confidence or neutral market regimes where the model suggests no clear edge.

The script allows for fine-tuning via the Minkowski distance parameter, enabling users to switch between different geometric calculations for pattern matching.

Details

The script follows a structured machine learning pipeline to ensure robust outputs:

• Feature Engineering: The model processes a multi-faceted feature set, including RSI variations, Price-to-MA deviations, and candlestick body dynamics.
• Normalization: All raw data is standardized into Z-scores. This ensures that features with different scales do not bias the KNN distance calculation.
• Dimensionality Reduction: To combat the "curse of dimensionality," the script compresses nine raw technical features into four principal components. This reduces noise and ensures that the nearest neighbors found are statistically significant.
• KNN Engine: The algorithm scans a historical window for the 'K' most similar patterns using the Minkowski Distance metric. A Gaussian Kernel is applied to weight the closest neighbors more heavily than distant ones.
• Target Objective: Instead of predicting absolute price targets, the model estimates latent momentum. This focuses on the "energy" of the market, identifying the probable strength and direction of the next move.

Settings

Machine Learning Engine

• K-Neighbors (K): Sets the number of historical neighbors to consider for the prediction.
• Learning Window Size: Defines the lookback period for the training data.
• Prediction Threshold: The confidence level (0.1 to 1.0) required to trigger a signal.
• Momentum Window: The lookback period used to label historical price direction as positive or negative.

Feature Engineering

• Feature MA Type: Selects the Moving Average type (SMA, EMA, HMA, etc.) used for deviation features.
• RSI Periods (Short, Mid, Long): Adjusts the lengths for the multi-period RSI features.
• MA Periods (Short, Mid, Long): Adjusts the lengths for the multi-period Price-to-MA features.
• Minkowski Parameter (p): Defines the distance metric; 1 is Manhattan distance, 2 is Euclidean distance.
• Shape Parameter: Controls the Gaussian weighting exponent for neighbor proximity.

Signal Filters

• Filter Condition Mode: Determines if signals must align with specific trend conditions (e.g., Price > Fast MA).
• Filter MA Type: Sets the MA type used specifically for the trend filter.
• Fast/Slow Filter Period: Defines the periods for the trend-filtering moving averages.

Dimensionality Reduction & Visuals

• Enable PCA Compression: Toggles the compression of features into Principal Components to reduce noise.
• Dynamic Bar Coloring: Enables/disables the confidence-based gradient coloring on price bars.

FAQ

How do I interpret the difference between "Major" and "Counter" signals? Major signals are those that align with the trend filter settings (e.g., a buy signal while price is above the EMA), whereas Counter signals trigger against the filtered trend and may carry higher risk.

What does the "Curse of Dimensionality" mean for this indicator? In machine learning, adding too many features can make data points appear equidistant, making KNN ineffective. This script uses PCA (Principal Component Analysis) to compress data, ensuring the "nearest neighbors" are truly relevant.

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