Predicting Customer Income With Data Mining Techniques
- 발행기관 경영전문대학원
- 지도교수 김진화
- 발행년도 2009
- 학위수여년월 2009. 8
- 학위명 석사
- 학과 경영전문대학원 MIS
- 실제URI http://www.dcollection.net/handler/sogang/000000045369
- 본문언어 영어
- 저작권 서강대학교의 논문은 저작권에 의해 보호받습니다
초록/요약
In this study customer income is predicted using U.S. census data employing data mining techniques. A comparative performance analysis of decision tree c5.0, back propagation neural networks, and linear regression models is conducted. This study also suggests the strategic implications of the results from the customer relations perspective for marketing strategy in practice and these results have various practical implications to retail banks with regards to decision making and strategic planning. This can also be applied for fraud detection of income gaps between predicted and actual incomes. The results show that decision tree c5.0 outperforms other models in predictability hence focusing more on decision tree and its advantages in lieu of prediction accuracy. In the mean time a set of decision rules are also extracted from the trained decision tree c5.0 in order to improve the clarity and explicability of the income prediction model.
more