Abstract
Semiconductor industry is capital intensive and competitive. Thus, it is important to utilize various resources efficiently to generate products for maintaining competitive advantages. Most of the existing studies on semiconductor performance evaluation focused on the company-to-company comparison. This study aimed to construct efficiency evaluation among the semiconductor fabrication facilities(fabs) since little research has been done on fab-to-fab efficiency evaluation (i.e. inter-fab evaluation.) In particular, the proposed model used multiple indices to measure inter-fab relative efficiency through data envelopment analysis. Managers can understand each decision making unit’s condition to set the improvement direction and make appropriate resource allocation decision through the inter-fab comparison. In addition, we applied backpropagation neural network to construct the inter-fab performance forecast model according to the characteristic of dynamic data update. Therefore, semiconductor companies can do real-time evaluation and make better strategies and resources planning. We used empirical data to validate the proposed model by analyzing the pattern between production scale and production efficiency. We gave improvement direction for each fab to provide references for the future fab construction. Managers can strengthen their advantage and improve their disadvantage for better inter-fab performance based on the advantage index and disadvantaged index for each fab. According to the inter-fab performance forecast results, managers can be alertness and adjust their resource allocation before the future efficiency go down. Thus, the results showed the practical viability of the proposed approach.