Font Size: a A A

Analysis And Forecast About Stock Market Trend Based On Support Vector Machine

Posted on:2021-05-04Degree:MasterType:Thesis
Country:ChinaCandidate:L ZhangFull Text:PDF
GTID:2428330614463730Subject:Applied statistics
Abstract/Summary:PDF Full Text Request
With the improvement of the economic level,the stock market is intensifying,and more and more people are investing in the research of stock prediction.However,there are many factors that affect stocks,including financial policies,market environment,economic cycles,and even human manipulation.How to seize the many influencing factors to find an effective method for stock forecasting is a hotspot for many experts and scholars.Among the many stock prediction methods,support vector machine is a good machine learning method.In this paper,the application of support vector machine in stock prediction is researched.First of all,a total of 24 stocks in various industries and industries were randomly selected from the Shanghai and Shenzhen Stock Series stocks,supplemented by the Shanghai Stock Index and the Shenzhen Stock Index,and promoted to the entire stock market by obtaining good prediction results on selected stocks.In terms of technology,this paper takes the model's prediction accuracy and average fitting deviation as the standard to measure the model's prediction performance,constructs appropriate feature variables through feature engineering,and uses the principal component analysis method to reduce the dimensionality of the input variables to eliminate the variables.Multi-collinearity can improve the speed of the model at the same time,build support vector classifiers and support vector regression machines to combine regression fitting and stock ups and downs classification to jointly predict stocks,and build stock investment strategy models based on support vector regression.While constructing the model,continue to explore the appropriate time sliding window,and try different parameter optimization methods including genetic algorithm and particle swarm optimization to optimize the prediction performance of the model as much as possible.While optimizing the model,mining search popularity and news sentiment,using natural language processing technology to try to find the correlation with the historical trend of the stock to build better emotional features,and further improve the prediction performance of the support vector machine model.This article also intends to build a multi-core support vector machine to further improve the effect of the model.Finally,through the division of "bull market" and "bear market",it attempts to build corresponding support vector machines at different stages to find a support vector machine model suitable for different scenarios,so as to achieve the right Better prediction of stocks.Empirical analysis found that through principal component analysis dimensionality reduction,the support vector machine model can greatly improve its running speed with a slight loss of prediction accuracy.At the same time,it is found that the best time window for the support vector machine to predict stocks is 3,The combination of support vector classification and regression can improve the prediction accuracy of the model by 3%-4%;in terms of parameter optimization,genetic algorithms and particle swarm optimization are all leading in the three directions of model iterations,running time and prediction accuracy Ordinary grid search method,at the same time,by constructing appropriate news sentiment characteristics,the prediction accuracy of the model is further improved;on the other hand,stock investment strategy can help investors choose the stock portfolio,which brings more Benefit,multi-core support vector machines can improve the performance of the model at an acceptable speed;the research also found that the support vector machine models built in the "bull market" and "bear market" are significantly better than the "consolidation period",especially in Under the blessing of search popularity index.Through the research in this paper,I hope that the continuously optimized support vector machine model can provide some guidance and reference for many investors and government departments.
Keywords/Search Tags:Support vector machine, principal component analysis, genetic algorithm, particle swarm algorithm, news emotion, stock forecast
PDF Full Text Request
Related items