Research On Machine Learning Based Multi-Factor Quantative Method And Its Application

With the development of software technology and artificial intelligence,more and more investors pay attention to quantitative trading.For stock investment,quantitative stock selection is the foundation,without good stock selection technology,the effect of quantitative investment will be greatly reduced.Multi-Factor quantization for stock selection is a method to quantify the various factors that affect the stock return into the indicators or factors that can be processed by computer or mathematical model,and use these indicators or factors to select stocks.This method can well exclude the influence of people's psychology and emotions to select stocks,and fully use the data to objectively and accurately analyze stocks.In recent years,with the development of the domestic financial market,more and more financial data have been accumulated.The quantitative stock selection method based on machine learning is gradually emerging.In this paper,we focus on the building of factors database,effective factors selection,stock selection model based on machine learning,the design and implementation of quantitative stock selection system to research.In this paper,we have used market data of stock trading and the financial statement data of listed companies to calculate 155 quantitative factors.These factors are classified and numbered,and build a basic quantitative factors database.Then,we use regression method and IC value analysis method to select some effective factors from the quantitative factors database,and use these factors as the basic data to train model.In the aspect of machine learning quantitative stock selection model design,this paper proposes a Two-Branch quantitative stock selection model,which breaks the barriers of technical analysis and fundamental analysis,and can use both technical data and fundamental data to conduct end-to-end analysis of stock selection.In the experimental part of this paper,the model is compared with the linear model based on the arbitrage pricing theory(APT)and the LSTM model,and the stock selection ability of the model is tested in the extreme market environment.The experimental results show that Two-Branch model can generate better returns than linear model and LSTM model,and the risk return performance in bear market and bull market is better than CSI 300.In the last part of the paper,a quantitative stock selection system based on machine learning model strategy is designed and implemented,which can test back and simulate the quantitative stock selection model based on machine learning.