Applying Data Mining Techniques to Evaluating Equity Investments

• Main Authors: Yi-rong Li, 李依蓉
• Other Authors: Yue-Shi Lee
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/31600049720867591325

碩士 === 銘傳大學 === 資訊工程學系碩士班 === 102 === High yield and high risk is the nature characteristic for most equity investments. While decisions are mainly made by human minds, it is crucial to apply decision models in investment evaluation period. However, due to the highly increased number of enterprises and low quality of their information, investors always found them jumping from one set of information silos into another. Thus, in this paper, we introduce data mining techniques into investment evaluation task in order to help those investors by quantitatively analyzing normal enterprises and choosing those which are most likely to be invested. In this we, we narrow down the list of enterprises that investors need to go through with, and make them able to focus on the further reach of these enterprises. Previous studies mainly focus on the categories of evaluation indexes, which are given by expert analysis. Our study extends it by specifying evaluation index with concrete measurements from real enterprises, and weighting them with feature selection techniques rather than subjective judgments. To handle the real data set, imbalanced data distribution is considered in this study. A precise and objective evaluation model is built then by using decision tree methodology from data mining, and different sampling method are used here, including SBC(Under-Sampling Based on Cluster) and Random Under-Sampling. Moreover, to handle the high-dimension problem properly, four feature selection techniques are taking involved: IV(Information Value), FS(Fisher Score), S2N(Signal to Noise ratio), and FAST(Feature Assessment by Sliding Threshold).