Optimization Of Stock Multifactor Model Based On Machine Learning

Nowadays,the financial market is booming,financial derivatives tools emerge in endlessly,the financial market is gradually improved,and financial supervision has been strengthened.The traditional stock fundamentals research has been difficult to match the complexity of the financial market,and portfolio strategy innovation has become the focus of academic research.The continuous development of machine learning algorithms and big data AI and the resulting changes provide theoretical and technical support for the establishment of quantitative stock selection models,and they use quantitative methods to build the portfolio strategy of stocks.The multi-factor stock selection model aims to find some factors most related to stock yield,assign different weights to the combination of these factors,then conduct comprehensive evaluation of individual stocks and screen out the high-quality stock construction portfolio,and find an investment hedging strategy with sustained Alpha returns.The construction of traditional multi-factor models is usually linear,while machine learning can express nonlinear factors,so as to more accurately grasp the market signal and obtain high and steady excess returns.In the selection of factors,this paper obtained CSI 300 stock data from 2006 to 2019,selected the factors that can significantly affect the stock yield,found factors IC,and finally selected the city net rate,net profit growth rate,turnover rate,flow rate,asset-liability ratio factors for model construction.According to the factor selection results,the historical data on the poly-width quantitative trading platform from 2018.01.01 to 2020.08.31.This paper analyzes the returns of three machine learning algorithms and linear stock selection strategies,Getting the policy yield of machine learning algorithm can exceed double the benchmark yield,The test indexes of annualized yield,Beta,Sharp ratio,maximum pullback rate and information ratio of several different stock selection strategies are analyzed here,The results show that machine learning based stock selection model compared with equiweight model,Sharp and information ratios are greater,This means in the assumption of the same risk,Machine learning share selection strategies can get more return on investment.Among them,the policy income of the support vector machine model exceeds four times the benchmark income,so this paper also combines the support vector machine stock selection strategy with the Alpha hedging risk control strategy,and finds that the systemic risk improves the portfolio yield while controlling the risk.Through machine learning method to build multi-factor stock selection model and combined with stock index futures hedging strategy,can provide investors with more excess returns from the stock market stock selection strategy,and through building machine learning stock quantitative investment strategy for excess returns,to provide theoretical support for risk control and stable returns of financial institutions.