Research On Stock Price Prediction Algorithm Based On Financial Index

The stock market has developed into an important part of China’s market economy for thirty years.More and more investors are paying attention to the stock market.How to predict the rise and fall of the stock price more accurately has become the most concerned topic among investors.Financial indexes,as one of the quarterly statements to be disclosed by listed companies,play a very important role in enterprise asset evaluation,enterprise credit evaluation and other aspects.They are also often used as an important indicator to judge a company’s financial status and to analyze the stock price of listed companies.Using artificial intelligence and data mining technology to find out the relationship between financial indexes and stock prices has become a popular research direction.This thesis used financial index and historical stock trading data to build the stock price fluctuation classification and regression prediction model used.The main innovative work of this thesis includes:(1)Construction of stock price forecast model based on financial indicators and random forest.This thesis used recursive feature elimination which based on decision tree to select the characteristics of the financial index data set.Financial indexes combined with historical stock trading data to build a random forest model.This thesis used genetic algorithm to optimize the key parameters of the random forest model,and realized the stock price rise and fall classification prediction.(2)Construction of stock price regression prediction model based on financial indicators and LSTM.This thesis used normalization and principal component analysis to preprocess the financial index data set and historical stock trading data set.This thesis used the LSTM as a basic regression prediction algorithm and genetic algorithm to optimize the time sequence of the LSTM,realized the stock price regression prediction model.(3)Experimental analysis and effect comparison.This thesis selected financial indicators and historical data of different listed companies in the financial industry to conduct experiments.By comparing with common stock price prediction algorithms such as decision tree,logistic regression,bayes classification,random forest and LSTM,it verified that the model of stock price rise and fall and the regression prediction model proposed in this thesis had good effect in the prediction of stock price fluctuation classification and regression prediction respectively.