Feature selection methods for financial engineering
Date of Issue2017
School of Electrical and Electronic Engineering
This paper is mainly focused on the academic goal. By reproducing the method in the existing published paper using the same data sets and recommend improvements on the procedure. Two main feature selection methods are used in this paper. One is the Classic Support Vector Machine regression and another one is an improved method Feature-weighted Support Vector Machine regression which was proposed by James N. K. Liu and Yanxing Hu. The improved method combines the Classic Support Vector Machine with the Grey correlation degree. Given different weight values to different features, the closer the relation between the feature to the target problem, the higher the weight value will be given. The processed data then goes into the Support Vector Machine regression training and the output model will be used for forecasting the stock daily close price. In this paper, the historical data and technical indicators of 7 stocks are downloaded from China Shenzhen A-share market. 5 stocks are used same as the reference data in the same period. Another 2 Growth Enterprise Market stocks are added to test generalization of the method. The result for this paper shows the Feature-weighted Support Vector Machine regression has better performance in forecasting the stock price.
DRNTU::Engineering::Electrical and electronic engineering
Final Year Project (FYP)
Nanyang Technological University