A hybrid fuzzy-neural approach to job completion time prediction in a semiconductor fabrication factory

Job completion time prediction is a critical task to a semiconductor fabrication factory. To further enhance the accuracy of job completion time prediction, the concept of input classification is applied to the back propagation network (BPN) approach in this study by pre-classifying a job with the fuzzy c-means (FCM) classifier before predicting the completion time with the BPN predictor. The data of jobs belonging to different categories are then learnt with different BPNs but with the same topology. After learning, these BPNs form a BPN ensemble that can be applied to predict the completion time of a new job. The output of the BPN ensemble determines the completion time forecast, and it is obtained by aggregating the outputs from the component BPNs with another BPN for nonlinear aggregation. Further, to improve the suitability of the FCM and BPN combination for the data, the prediction error by the BPN is also fed back to adjust the classification result by the FCM. On the other hand, the future release plan of the factory, which is influential but has been ignored in traditional approaches, is also incorporated in the proposed methodology with up to eight future discounted workloads. Production simulation is also applied in this study to generate test data. According to experimental results, the prediction accuracy of the proposed methodology was significantly better than those of some existing approaches.