A MANUFACTURING FAILURE ROOT CAUSE ANALYSIS IN IMBALANCE DATA SET USING PCA WEIGHTED ASSOCIATION RULE MINING

Abstract

Root cause analysis is key issue for manufacturing processes. It has been a very challenging problem due to the increasing level of complexity and huge number of operational aspects in manufacturing systems. Association rule mining (ARM) which aids in root cause analysis was introduced to extract interesting correlations, frequent patterns, associations or casual structures among items in the transactional database. Although ARM was proven outstanding in many application domains, not many researches were focusing on solving rare items problem in imbalance dataset. The existence of imbalanced dataset in manufacturing environment make the classical ARM fails to extract interesting pattern in an efficient way. Weighted association rule mining (WARM) overcomes the rare items problem by assigning weights to items. The goal of using weighted support is to make use of the weight in the mining process and priorities the selection of the selection of targeted itemsets according to their significance, rather than frequency alone. However, the development of a suitable weight assignment scheme has been an important issue. In this research, we proposed principal component analysis (PCA) to automate the weight in WARM. The result shows that PCA-WARM is capable in capturing pattern from the data of industrial process. These patterns are proven able to explain industrial failure.