Stock Price Prediction Based On Optimized GAM Model

An increasing number of people are attempting to excavate the hidden and valuable information from the vast amount of stock data.However,since the market is subject to the influence of a variety of factors,the stock price is unpredictable with a non-linear and non-stable trend.Great research effort has been input to explore how to predict the trend and the fluctuation range of the stock price as well as how to select a appropriate prediction method.This paper focuses on stock price prediction and stock investment strategy based on massive data.The main content is summarized below:Firstly,through the quantification of the factors that affect the stock market,this paper proposes a prediction method based on the optimized generalized additive model.Specifically,the Fourier series and the logistics block growth model are used to quantify the predictive function,transforming the non-linear problem into a linear one.This algorithm is adopted to predict the closing price of international and domestic stocks.Afterward,the model is trained through a back-fitting algorithm and a spline smoothing algorithm.In order to achieve better prediction,we get trend lines of the stock price with minimum smoothing errors and a good fitting effect based on changing point prediction and the regressive algorithm of OLS.From the empirical analysis of a large amount of data,some interesting phenomena are found,such as “End Effect”,“Black Thursday” and “Predictive method from small to large”.The results show that the predictive accuracy of the optimized generalized additive model is up to 89%,11%higher than that of RBN,SVM,SSA-SVM and other models.Secondly,in order to help novice investors to identify the fluctuation trend of the stock market,this paper further puts forward an S-DT investment strategy based on the optimized decision tree model.In order to improve the predictive accuracy,the indexes highly correlated with the trend of closing price are selected as the input features of the decision tree model with such methods as synergy factor,information entropy and grey association rules.The empirical data analysis shows that the predictive accuracy of the S-DT investment strategy reaches 75.5%,which is 3.9% higher than that of GRA-DT,ANN,DM and other models.To sum up,the predictive model of the stock price is quantified and optimized through various factors to improve the predictive accuracy of the stock price.And thedecision tree model is optimized with several indicators closely related to the trend of the stock price.On this basis,a stock investment strategy is put forward.Thus,the predictive model of the stock price provides a reference for predicting the fluctuation trend of the stock price.In addition,novice investors can also make predictions based on stock investment strategies,thereby maintaining and increasing their value.