Overview

This is an experimental quantitative trading strategy that combines moving average indicators and kNN machine learning algorithms to generate trading signals. It uses the crossovers of two VWMA lines with different periods to determine the trend direction, and uses MFI and ADX indicators to filter the signals through kNN algorithm to improve the reliability of the signals.

Strategy Principles

The core indicators of this strategy are two VWMA lines with different parameters, namely fast line and slow line. When the fast line crosses above the slow line, a buy signal is generated. When the fast line crosses below the slow line, a sell signal is generated. In addition, this strategy introduces two auxiliary indicators, MFI and ADX, to judge the reliability of the current signal under the current market conditions through the kNN classification algorithm.

The idea behind kNN algorithm is to compare new data with historical data to determine the results corresponding to the k most similar historical data, and categorize based on the majority vote of these k historical results. This strategy uses MFI and ADX as two input parameters of the kNN algorithm to determine the historical price movement (uptrend or downtrend) under this combination of indicators, thereby filtering the current signal to improve the signal quality.

Advantages

• Utilize the trend following capability of VWMA and generate trading signals through moving average crossovers
• Apply MFI and ADX indicators for multi-dimensional feature extraction to assist in determining trend direction
• Leverage kNN machine learning algorithm to dynamically optimize and filter trading signals
• Experimental strategy with large room for improvement through more data verification and optimization

Risks and Mitigations

• VWMA lines tend to lag
• MFI and ADX have some lagging, which may misjudge market conditions
• kNN algorithm parameters (e.g. k value) can have significant impact on results
• Experimental strategy may underperform in live trading

Mitigations:

• Adjust MA parameters to reduce lag
• Improve indicators to judge trends more accurately
• Optimize kNN parameters to improve fitness
• Verify strategy via backtest and paper trading

Optimization Directions

There is large room for optimizing this strategy:

• Add more MA indicators to construct MA combinations
• Try different auxiliary indicators like MACD, KDJ etc
• Improve kNN algorithm e.g. using different distance metrics
• Try other machine learning algorithms like SVM, Random Forest etc
• Parameter tuning to find optimal parameter sets

Introducing more indicators and machine learning algorithms may further improve the stability and profitability of the strategy.

Summary

This is an experimental quantitative trading strategy based on VWMA indicators and kNN machine learning algorithms. It has the advantage of strong trend following capability while filtering signals via machine learning. The strategy has large room for expansion by introducing more features and optimization algorithms for better results. But as a novel strategy there are also risks that require further verification and improvement. Overall this strategy has great innovation potential.