英语论文网

International Journal of Information Technology & Decision Making
World Scientific Publishing Company
A DECISION TREE-BASED CLASSIFICATION APPROACH TO RULE EXTRACTION FOR SECURITY ANALYSIS
Department of Computer Science
Southern Illinois University Carbondale
Mailcode 4511, Carbondale, IL, 62901-4511, USA
∗rahimi@cs.siu.edu
Stock selection rules are extensively utilized as the guideline to construct high performancestock portfolios. However, the predictive performance of the rules developed bysome economic experts in the past has decreased dramatically for the current stockmarket. In this paper, C4.5 decision tree classification method was adopted to construct
a model for stock prediction based on the fundamental stock data, from which a set ofstock selection rules was derived. The experimental results showed that the generatedrules have exceptional predictive performance. Moreover, it also demonstrated that the
C4.5 decision tree classification model can work efficiently on the high noise stock datadomain.

Keywords: Stock selection rules; stock prediction model; decision tree; data mining;

C4.5 decision tree algorithm.
1. Introduction
The stock market is a complex system affected by various factors which make 代写留学生论文it veryhard to be predicted. Research has been conducted to predict the future behaviorof the stock market using various techniques and approaches for which computertechnology has played a very important role. Today, there are various stock analyzerprograms available as commercial products or research projects. They range frompure mathematical models, databases, and expert systems to neural networks andfuzzy systems. These software packages are based on either a technical analysisapproach or a fundamental analysis approach. Stock selection rules are the truespirit of most of the fundamental analysis based portfolio selection systems.
In the real world, when an investor selects stocks to construct an optimal portfolio,some rules of thumb are usually used to evaluate stocks. For instance, usuallystocks with a stable earning record, low price-to-earning ratio, low debt, and/orhigh dividend yield are selected. Without any doubts, there are some principal
rules associated with choosing the right stocks, although the rules may differ fromone era of economics to the other. Finance experts have set up stock selection rulesmainly based on some finance and economic theories such as Peter Lynch investmentrules, Warren Buffet investment rules, Benjamin Graham investment rules,Philip Fisher investment rules, and T. Rowe Price investment rules.1 Many of theabove rules have had good performances during some period of time in the past.For example, Graham’s ten rules had their best performance before 1976. However,with the fast development of the stock market and the globalization of theworld economy, present stock markets have changed dramatically as compared tothe market prior to 1980. The applicability of many existing rules is unsatisfactoryand their performance has decreased substantially. Therefore, the main objectiveof this study is to produce a new set of rules, which are to be used to select stocks
with possible high return.In recent years, some data mining based techniques such as classification, ruleinduction, etc., have been utilized to analyze the stock market which have led tothe development of promising stock market prediction models.2–5 John and Miller2introduce the本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。