Affiliations: [a] Department of Information Technology, Sri Krishna College of Engineering and Technology, Coimbatore, India | [b] Department of Computer Science and Engineering, Government College of Technology, Coimbatore, India
Correspondence:
[*]
Corresponding author: S. Shankar, Associate Professor, Department of Information Technology, Sri Krishna College of Engineering and Technology, Coimbatore, India. E-mail: shanx80@yahoo.in.
Abstract: Utility based data mining has been emerging as a significant learning tool for formulating business practices from the economic perspective. The recent economic growth in financial world could be greatly accredited to the Stock markets. Many researchers have made valiant attempts in data mining to devise an efficient system for stock market movement analysis. Still, there have not been many systems that could formulate association rules for stock markets based on economic importance i.e. utility. So, when utility based data mining is employed in stock market prediction, it would devise association rules that would lead to high return on investments. Hereby, we propose a new analysis which deploys the Utility Based Data mining to generate utility emphasized trading rules. The proposed utility based analysis is composed of a pre-analysis and a core analysis. The pre-analysis utilizes four powerful technical indicators to interpret the raw historical data and Association Rule Mining to generate frequency-based trading rules. In the core analysis, the utility based preliminary rules are generated by using Genetic Algorithm and then a weightage based analysis is performed for extracting better utility-emphasized rules. Unlike the conventional utility based data mining, the obtained trading rules are optimal and utility efficient because of the introduction of weightage based analysis. The proficiency of the obtained rules are compared against the conventional utility-emphasized and frequency-based (non-utility emphasized) trading rules using frequency, utility and weightage values in diverse stock market datasets.
Keywords: Utility based data mining, utility analysis, utility emphasized trading rules, Association Rule Mining (ARM), Genetic Algorithm (GA)
