Predictability Of Chinese Stock Price Based On Machine Learning

With the enhancement of computer hardware and software functions and the rapid development of theories and technologies such as artificial intelligence and data mining,the research and application of machine learning has set off a frenzy.Nowadays machine learning has been widely used in all aspects of human life.Among them,the application of machine learning algorithm to stock price prediction has become one of the research hotspots in the financial industry.Traditional methods of studying stock prices mainly use technical index analysis or time series analysis.However,the traditional analysis methods have great limitations in dealing with the non-linear data of stock prices.But machine learning method can deal with the randomness,chaos and non-linearity of stock market well.Therefore,based on previous research,this paper uses machine learning algorithms to empirically analyze the predictability of China’s stock price.In this paper,three comruonly used machine learning algorithms are selected to study the predictability of China’s stock price.They are artificial neural network(ANN),support vector machine(SVM)and random forest(RF).The stocks of Shanghai and Shenzhen 300 index are selected as the research object,and the technical index of stock price is taken as the input variable of the model.The stock yield after discretization is taken as the predictive variable of the model.Adjusting the parameters of the model by grid search and cross validation.Comparing the AUC mean of the optimal model of each algorithm,it is found that the prediction results of different algorithms for the same data set are quite different,among which the random forest algorithm has obvious advantages on the data set in this paper.Then use the most suitable algorithm to predict the stock price movement trend of each stock from 2014 to 2018,and the forecasting performance of the model is comprehensively evaluated by using multiple evaluation indicators.The forecasting results show that the average value of the evaluation indicators of all years is above 0.5,and the maximum value of the evaluation indicators of the model is above 0.7,which shows that the model built in this paper has good prediction performance,which shows that it is feasible to use machine learning algorithm to predict the trend of stock price change in China.Then,based on the forecast results of 2017 and 2018,this paper constructs the portfolio,which gains excess returns in the market compared with the actual rise and fall of Shanghai and Shenzhen 300 index.It further demonstrates the effectiveness and practicability of the machine learning model in stock price prediction.In addition,by comparing the predictive evaluation indicators of different stocks,it is found that the same model has different predictive effects on different stocks.Therefore,this paper studies the correlation between the characteristics of each stock and the predictability of its stock price.Through industry analysis,it is found that the predictability of each stock in three industries,namely,banking,oil and gas,electricity,is the strongest.Using this discovery,investors can pay more attention to stocks in these industries when they choose stocks.By analyzing the correlation between other characteristics of stocks and the predictability of stock prices,we find that the bigger the market value,the stronger the predictability and the smaller the earnings per share,the stronger the predictability.This analysis can provide a certain reference direction for investors to choose stocks.