Research On Quantitative Investment Strategy Based On Voting Model

With the advent of the era of financial big data,financial data has been greatly enriched in type and quantity,and machine learning has become the frontier of application in the financial field.According to the characteristics of the stock market,retail investors gradually flow to institutions and invest Behavior has also changed from active to passive,especially in building quantitative investment strategies,under which quantitative investment has become a hot spot in the investment community.The machine learning voting model is a classic algorithm in machine science,which can effectively model based on the massive and high-dimensional characteristics of stock data.Based on the voting model,this paper discusses the feature selection and data modeling in quantitative stock selection.In this paper,the clustering method is used to derive the company’s characteristic variables in my country’s stock market,and then based on the semi-supervised learning method,the missing values of all the company’s characteristic variables are filled,and finally all the obtained factors are used to construct multi-factors through machine learning technology.A stock return prediction model,and then using real data to build a portfolio based on this model to backtest the effectiveness of the strategy,found that the annualized return of the long-short group in the portfolio reached 13.8%.This paper also further analyzes the source of income of the machine learning model from the perspective of mispricing,and empirically proves that our source of income can be explained by the theory of financial mispricing.With the enrichment of company characteristic variables and the timeliness of the market,the combination of machine learning technology and factor research is the general trend.Building a multi-factor quantitative investment strategy based on machine learning technology can more efficiently and reasonably predict asset pricing.The management of the Chinese market and the effective omission of information.