Clustering with removing outliers - a case of an IC packaging company

摘要

Companies that produce semiconductors are exposed to greater challenges because of complex manufacturing systems used and competitive business environments encountered. Improving the machine performance has been and will remain to be a critical role in deciding the success of a company. In this paper, we apply the clustering techniques to partition the wire bond machines in a packaging company into groups so that management levels can improve the productivity based on the clustering results. Moreover, to gain an insight into how outliers affect the clustering results, we remove those outliers based on the visual results of self-organizing map (SOM) algorithm. Comparing this result with the one with outliers, we found that removing outliers leads to better clustering based on the Wilk's lambda value.