Prediction Of Stock Price Based On LS-SVM

With the progress of society and the improvement of the level of science and technology,important changes have also taken place in the data generated in real life.This change is mainly reflected in the amount of data,and the amount of data generated today is getting larger and larger.Such changes allow us to have a clearer understanding of the changing rules behind the data,but due to the increase in the amount of data,data mining work is made.more difficult.The data generated in real life is often with time attributes.Therefore,the data mining work in recent years has been slowly changing from static data to analysis of time series data.For example,in the financial industry,quantitative analysis of the financial industry using data mining techniques is produced.It mainly uses mathematical and modern statistical methods to extract useful information from the massive historical data of the financial market and hopes to use this information.Get better return on investment.This article uses stock data of Meidi Group from September 2013 to April 2018.And using the traditional time series analysis method to analyze the data,it is concluded that the closing price sequence in the stock data is a non-stationary time series.Then the ARIMA model is used to analyze the unsteady time series data and further obtain the time.The sequence is a random walk process,so that the stock closing price is randomly fluctuating based on the closing price of the stock the day before.Then choose a multiple linear regression model to model the closing price of the stock,and explain the relationship between the stock closing price on the second day and the stock closing price,open price,highest price,and lowest price on the first day.It is concluded that the order of the closing price of the second day of the stock from high to low is the closing price,the highest price,and the lowest price of the previous day.Finally,using the machine learning algorithm LS-SVM to predict the closing price of the stock,and selecting the method of segmented prediction to improve the effect of the model,so as to obtain the best model,and compare the obtained model with the previous multiple linear regression model.In the far-reaching linear regression model,the mean square loss on the test set is 0.6053,while in the model based on LS-SVM,the mean square loss is 0.4982.This shows that the prediction model based on LS-SVM is slightly better than the multiple linear regression model...