Panel manufacturers’ offering of high-quality products to customers is the most basic service. Thus, when customers obtain bad products, they will ask manufacturers to find the causes of abnormal products returned in shorter time and have the solutions. “Analysis” is adopted to find the causes of abnormal product quality. Although various quality analysis techniques, such as Fishbone Diagram and 5Why Analysis, are commonly applied to upgrading of quality in manufacturing industry, they do not significantly enhance the efficiency of “analysis” which relies on cumulative experiences. However, people’s inheritance of experience will be neglected by the original cumulative analytical knowledge and it will not enhance analytical efficiency. Data Mining means to find specific models or related information from great amount of data by computers. The information helps decision makers to have significant judgment. Thus, efficiency of analysis on bad products can be reinforced by this method. This study focuses on data of automobile panel “Line Defect” returned of case company and by Data Mining, it tries to find if the technique can effectively help enhance efficiency of analysis and if the new model can be applied to analysis of returned products. The conclusions are below: 1.Line Defect analytical rules by Data Mining are reliable to judge the units of abnormality. Thus, efficiency of analysis on returned products can be enhanced by this model. 2.Before Data Mining analysis, it must recognize the essence of the problems and the technique applied in order to find the information needed for decision making.
