Decision Tree and Data Mining Models for Product Upgrade Decision: An Empirical Study of Uninterruptible Power Supply Company in Taiwan

• Main Authors: Kai-WeiChen, 陳凱緯
• Other Authors: Chia-Yen Lee
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/87w5t4

碩士 === 國立成功大學 === 工程管理碩士在職專班 === 103 === Due to global warming, human put more focus on energy consumption and use of stable power quality. Nowadays, the “Uninterruptible Power Supply” are gradually becoming an essential part of the important equipment to satisfy the customers’ development request. The companies of UPS in Taiwan need to fully understand the characteristics of the industry to create the value and keep enterprise competitive advantage. It can reduce the risk of product failure and enhance service quality by the product upgrade decision and maintenance in field. In this study, we used the influence diagram and decision tree analysis to analyze the upgrade decision related to the UPS. The first step is to use influence diagrams and decision tree analysis for active upgrade, passive upgrade and no upgrade within or without warranty decision. The result showed that, combined with decision tree analysis and expectations of monetary value method to identify decision criteria can enhance the decision quality. With a sensitivity analysis to identify active and non-active upgrade the transition probability of the decision to change. We used of decision tree and logistic regression algorithm further application by data mining prediction model. The data mining of decision tree model created UPS failure rule and found two key factors the function block and the operation time, as same as we found by the logistic regression model. We knew a comparison of classification accuracy between decision tree and logistic regression. The result will provide UPS enterprises with the follow-up reference of product upgrade decision. It will make win-win decision not only for customer satisfaction but also for from profit maximization.