Using Discretization of Numeric Attributes to Compare the Changes in Performance of C4.5 and CART algorithms

• 발행기관: 한국유통과학회
• 저자명: Yong-Gyu Jung Jun Heo Kyu-Ho Kim
• 간행물 정보: 『KODISA ICBE (International Conference on Business and Economics)』2013 International Conference on Business and Economics (ICBE 2013), 353~358쪽, 전체 6쪽
• 주제분류: 경제경영 > 경제학
• 파일형태: PDF
• 발행일자: 2013.07.13

영문 초록

Purpose - Actual learning techniques applied to data mining problems can substantially improve the success of these topics. Mechanical study of how a vast array tested. Such as decision trees, decision rules, schema based on linear models, for example, numerical prediction techniques, clustering algorithms and Bayesian network.. Research design, data, methodology - Actual data mining problem are applied to both of these outstanding seems to be a powerful technique. Useful information can be applied to real-world navigation and finding the relationships that exist within these data using large amounts of data from a well-established statistical techniques and machine learning techniques, patterns, rules, etc. They are modeled by ‘Actionable Data Mining Information’ to extract a series of process can be defined. Results - Data mining has the effect of the analyst did not intend to pull hypotheses output to generate the dictionary of useful information. These useful information in order to generate meaningful information from the data by using a variety of data mining techniques to find them. Using some techniques for data collection, screening and must decide whether to create a model. Conclusions - Therefore, in this paper, which is one of data mining techniques that can be used as algorithm C4.5 and CART algorithms, performance was compared against the same data. In the future, more precise experiments to establish realistic and similar environmental factors, the performance will be evaluated.

목차

Abstract

1. Introduction

2. Literature Review

3. Research Methodology

4. Results

5. Discussions and Policy Implications

6. Conclusions and Research Limitations

References

키워드

k-NN Instance-based Learning Decision Tree Feature Selection Symmetric Uncertainty Discretizing Breakpoints