Establish a cluster based evolutionary adaptive Weighted Fuzzy CBR for PCB sales forecasting

Reliable prediction of sales can improve the quality of business strategy. Case-Based Reasoning (CBR), one of the well known Artificial Intelligence (AI) techniques, has already proven its effectiveness in numerous studies. However, due to the uncertainties in knowledge representation, attribute description, and similarity measures in CBR, it's very difficult to find the similar cases from case bases. In order to deal with this problem, fuzzy theories have been incorporated into CBR allowing for more flexible and accurate models. This research develops a hybrid model by integrating Self Organization Map (SOM) neural network for data clustering, Genetic Algorithms (GAs) for parameters optimization and Weighted Fuzzy CBR (WFCBR) as main forecasting model to forecast the future sales in a printed circuit board (PCB) factory. This hybrid model encompasses two novel concepts: 1. Clustering WFCBR into different clusters by adopting SOM, thus the interaction between WFCBR is reduced and a higher accurate prediction model can be established. 2. Evolving WFCBR by optimizing the variables weights and fuzzy term numbers of the inputs and outputs, thus the prediction accuracy of the WFCBR can be further improved. Numerical data of various affecting factors and actual demand of 5 years of the PCB factory are collected and fed into the hybrid model for future monthly sales forecasting. Experimental results show the forecasting accuracy is obtained by the proposed hybrid model and it is superiors to the other comparing methods.