A Stock Prediction Model Based On The Cart And Boosting Algorithm

The stock market is an important embodiment of the market economy,and it also reflects the status of China's economic development.It plays a crucial role in the analysis of economic development trends.With the development of the stock market,there are more and more people who choose to invest in stocks as profitable projects.However,how to choose stocks and how to choose the corresponding listed company for investment to obtain maximum returns has always been a problem.Therefore,it is of great significance to analyze and predict the trends of various stocks on the stock market.However,stock data is a large and chaotic complex system.It is difficult to predict and analyze it using traditional methods.Data mining technology provides a good solution to extract valuable data that may be valuable for trend forecasting from a large number of disorganized data.For the current stock prediction model is not high enough accuracy,there are problems such as over-fitting or under-fitting.Based on the analysis of existing stock prediction methods,a stock prediction method based on Cart decision tree and Boosting method is proposed.Based on the vertical correlation of the data,this method adds the “average price in the past ten days” and the “transfer rate” in two vertical directions based on five commonly used opening prices,closing prices,trading volumes,the highest price of the day,and the lowest price of the day.Based on the Cart decision tree method,Boosting cascades multiple decision trees to solve the fitting problem.This article selects a listed company in the instrumentation field of the A-share market as sample data,selects stock transaction data of a company in the field for one year as an input variable,uses a C5.0 decision tree,a Cart decision tree five indicator,and a Cart decision tree combines Boosting Algorithm Seven indicators Three kinds of machine learning algorithms respectively establish prediction models for comparative analysis.Select the data of a company's A-meter instrument as sample data and use this predictive model to train.Among them,80% of the data was randomly selected as the training set,and 20% of the data was used as the test set to test the validity of the model.Through comparison of the prediction accuracy of the three classification models,it is found that comparing with the existing prediction models,the accuracy of the Cart model combined with the Boosting algorithm plus seven input variables is improved,and the predicted mean square error is decreased by 0.22.Using this model to analyze the characteristics of stocks in a specific field can help investors to make investment decisions in stocks.