Optimization of ETF Fund Selection Strategy Based on Machine Learning Scoring Model

Abstract

This study aims to optimize the ETF fund selection strategy. By combining the time-series momentum strategy and the machine learning scoring model, a systematic analysis and empirical research on the main stock ETFs and cross-border ETFs in the Chinese market are conducted. We constructed a combination strategy that combines momentum indicators and machine learning scoring, and conducted a backtest. The results show that the combination strategy is significantly better than the momentum strategy or machine learning model alone and the benchmark index in terms of annualized return, maximum drawdown and Sharpe ratio. By selecting momentum indicators (such as simple momentum, exponential moving average, relative strength index) and machine learning models (such as logistic regression, support vector machine, random forest), we verified the effectiveness of these methods in ETF fund selection. The backtest results show that combining different methods can effectively improve the performance of ETF fund selection strategies and achieve better investment results. This paper further analyzes the investment results of each strategy, and evaluates the stability and risk control ability of the strategy through cross-validation and backtesting methods. The results show that the combination strategy performs well in terms of return, risk control and risk-adjusted returns, providing investors with a scientific and effective investment tool.