Research And Improvement Of Quantitative Investment Strategies Based On Machine Learning And Volatility Aggregation Analysis

This paper mainly studies the application and advantages of machine learning in quantitative investment.We verified the feasibility that machine learning algorithm can improve the optimization of some quantitative investment strategies.Quantitative investment and machine learning have adequately demonstrated their unique advantages and charms in this big data era and they become even more powerful by combination.Machine learning not only help to overcome the disadvantages of quantitative investment,but also provides new ideas and directions for its development.Markets are constantly changing,so it makes sense to keep exploring which factors are more effective in quantitative investment.This paper mainly studies whether machine learning algorithm can be applied in factor composition,and whether its effect is better than the traditional method.In the empirical study,two indexes,prediction accuracy and AUC,are selected to evaluate three machine learning algorithms and linear logistic regression model.Evaluation indexes such as annualized rate of return,Sharpe ratio,information ratio,maximum drawdown and volatility are selected to compare and analyze the results of factors' layered back test.This paper points out that machine learning algorithm has guiding significance for quantitative investment and stock prediction,providing investors a feasible plan for decision-making.In addition,this paper introduces the phenomenon of volatility clustering and the financial time series model in detail,and applies the GARCH model to fit the market price volatility of CSI 300 index,verifying that the market price return rate of CSI 300 index has significant volatility clustering.Then,this paper creatively adds the volatility clustering as a "factor" to the machine learning part of the strategy,and compares and analyzes the back test results in detail by taking the initial strategy as a comparison.We concluded that the practical application of volatility clustering is worth studying,which can improve the performance and robustness of quantitative strategy,and also provide a new idea for the application and development of volatility clustering.Our research and improvement on XGBoost algorithm and volatility clustering provide a new perspectives for quantitative investors in CSI 300,which also provide certain reference significance for machine learning methods and volatility clustering's application in quantitative investment.