J Intell Manuf (2011) 22:585–595 DOI 10.1007/s10845-009-0321-7 Classification knowledge discovery in mold tooling test using decision tree algorithm Duen-Yian Yeh · Ching-Hsue Cheng · Shih-Chuan Hsiao Received: 28 April 2009 / Accepted: 21 September 2009 / Published online: 7 October 2009 © Springer Science+Business Media, LLC 2009 Abstract The scale of Taiwan’s mold industry was ranked the sixth in the world. But, under the global competitive pres- sure, Taiwan has lost its competitive advantage gradually. The only chance of Taiwan’s mold industry lies in improv- ing the competitive abilities in product research, develop- ment and design. In mold manufacturing cycle, mold tooling test plays a very important role at accelerating the speed of production. An experienced engineer can minimize the error rate of mold tooling test according to his rich experiences in parameter adjustment. However, this experience is mostly implicit without theoretical basis and its knowledge is diffi- cult to be transmitted. Benefiting from the well development of data mining technologies, this study aimed at construct- ing an intelligent classification knowledge discovery system for mold tooling test based on decision tree algorithm, so as to explore and accumulate the experimental knowledge for the use of Taiwan’s mold industry. This study took the only high-alloy steel manufacturer in Taiwan for case study, and performed system validation with 66 record data. The results showed the accuracy rates of prediction of training data and testing data are 97.6 and 86.9%, respectively. In addition, this study explored two classification knowledge rules and proposed concrete proposals for tooling test param- eter adjustment. Moreover, this study provided two ways, rule verification and effectiveness comparison of four mining algorithms, to conduct model verification. The experimental D.-Y. Yeh Department of Information Management, Transworld Institute of Technology, Yunlin, Taiwan e-mail: yeh@tit.edu.tw C.-H. Cheng (B ) · S.-C. Hsiao Department of Information Management, National Yunlin University of Science & Technology, 123 University Road, Section 3, Douliou, Yunlin, Taiwan e-mail: chcheng@mis4k.mis.yuntech.edu.tw results showed the decision tree algorithm has an excellent discriminatory power of classification and is able to provide clear and simple reference rules for decisions. Keywords Mold tooling test · Knowledge discovery · Data mining · Decision tree · Mold industry Introduction In Taiwan, the mold industry is known as the “mother of industry,” which is the foundation of mass production of products. Molds are the best tools for product standardiza- tion, mass production and cost cutting of either simplest com- modities or highly precise electronic parts. Figure 1 shows the import and export values statistics of the mold industry in Taiwan from 1998 to 2007. The annual gross values in the recent 4 years were among NT$55–60 billions, which ranked Taiwan the sixth in the world. In detail, the import and export values distribution between Taiwan and various countries in 2007 is shown in Table 1. As seen, Mainland China was the largest market to Taiwan, which accounts for 30.99% of the export value and 44.29% of the import value (Taiwan Mold and Die Industry Association 2009). In the application market, Taiwan gives priority to molds for electronic communication products. The characteristics of electronic communication industry include shorter product life cycle, shorter time to Market, and more product types but small quantity. The electronic communication industry needs to improve the design capacities of products and also expe- dite the development speed of products, in order to respond to severe challenges. In fact, in the cycle of product devel- opment, the design and manufacturing capacities of molds are key elements. In terms of economy and technology, the relative advantages of Taiwan include fast pre-production, 123