Font Size: a A A

Stock Selection Based On Multi-Class SVMs

Posted on:2011-01-08Degree:MasterType:Thesis
Country:ChinaCandidate:J H ChenFull Text:PDF
GTID:2219330362456827Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
As for us, the stock market is a complex system which is full of different kinds of information. Usually investors have to face with an enormous amount of stocks in the market. So, how to implement the selection of stocks to invest is an important process for them. From the perspective of data mining, the problem of stock-selection aims to identify stocks which can outperform the market (achieve exceptional returns) in the following year. At the same time, it's also a classification problem that detects the pattern or mappings between stock indicators and returns, from which the process of stocks selection can be achieved.The study combines the features and the actual operation situation of the domestic stock market, and then conducts the application of classification based on SVM in stocks selection. In order to reduce the non-linear classifier's complexity, the paper uses PCA technique on the financial indicators data of listed companies to complete attribute reduction without compromising classification accuracy. This method can achieve a few principal components which can retain the information of original data as much as possible. So, it can avoid the impact on classification's complexity due to that some properties of the input matrix are highly correlated, which can also improve the model generalization among different data sets. Investing by the stocks selected by the multi-class SVMs raised in this paper with the equally weighted portfolio, no matter when it is in a bull or bear market, investors can also get exceptional returns. Therefore it can be obtained that this study is a good alternative for investment.
Keywords/Search Tags:Support Vector Machine(SVM), Classification, Stock Selection, Principal Component Analysis (PCA)
PDF Full Text Request
Related items