VWMA with kNN Machine Learning: MFI/ADX

The VWMA with kNN Machine Learning: MFI/ADX indicator utilizes a volume-weighted moving average crossover system combined with a k-Nearest Neighbors classification algorithm to filter trade signals based on trend strength and money flow.

Usage

The Usage section describes how the script can be used to identify trend reversals and filter entries using machine learning. The tool generates signals based on the crossover of a "Fast" and "Slow" Volume-Weighted Moving Average (VWMA).

• Signal Generation: A bullish signal is triggered when the Fast VWMA crosses above the Slow VWMA, while a bearish signal is triggered when the Fast VWMA crosses below the Slow VWMA.
• kNN Filtering: If enabled, the kNN filter evaluates the current market state by looking at historical data points where similar ADX (trend strength) and MFI (volume-weighted momentum) values occurred. It then "votes" on whether the current crossover is likely to be successful.
• Background Visualization: Users can enable a background highlight to visualize the kNN model's current sentiment; green backgrounds indicate a bullish classification, while red indicates a bearish classification.

Details

This script implements a k-Nearest Neighbors (kNN) classification algorithm, which is a non-parametric supervised learning method.

1. Feature Space: The model uses two features—the Average Directional Index (ADX) and the Money Flow Index (MFI)—to create a two-dimensional coordinate system.
2. Data Storage: Every time a VWMA crossover occurs, the script stores the current ADX and MFI values along with the subsequent price action (whether the price moved up or down).
3. Classification: When a new crossover occurs, the algorithm calculates the Euclidean distance between the current ADX/MFI values and all stored historical points. It identifies the "k" closest neighbors.
4. Voting: The script aggregates the results of these neighbors. If the majority of similar historical instances resulted in positive price movement, the filter allows a long entry.
5. VWMA: Unlike standard simple moving averages, the VWMA gives more weight to bars with higher volume, providing a price average that is more representative of where the bulk of trading activity occurred.

Settings

VWMA Settings

• Source: Determines the price data used for the moving average calculations (e.g., Close, Open, OHLC4).
• Fast Length: The period for the short-term volume-weighted moving average.
• Slow Length: The period for the long-term volume-weighted moving average.

Filter Settings

• Apply kNN filter: Toggles the machine learning filter on or off.
• Filter Length: The lookback period used to calculate the ADX and MFI features.
• Filter Smoothing: The smoothing factor applied to the ADX calculation.

kNN Settings

• kNN nearest neighbors (k): The number of historical data points (neighbors) the algorithm considers when making a prediction.
• kNN minimum difference: The required threshold for the voting majority to trigger a trade signal.
• Draw background: Enables or disables the background coloring based on the kNN model's prediction.

FAQ

How do the ADX and MFI help the machine learning model? The ADX measures the intensity of a trend regardless of direction, while the MFI measures buying and selling pressure using volume. By mapping these, the kNN algorithm can identify if crossovers occurring in specific "volatility and flow" environments have historically been profitable.

Why does the model need a "kNN minimum difference"? This setting acts as a confidence threshold. Instead of a simple majority (e.g., 51%), a higher minimum difference ensures that a significant majority of historical neighbors agree on the direction before a signal is validated.

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