A novel auto-weighting deep-learning fuzzy collaborative intelligence approach

Experts often have unequal authority levels in organizations. However, this has rarely been considered reasonably in solving problems in the manufacturing system. In addition, a fuzzy collaborative estimation method can be more flexible and effective if experts have unequal authority levels. We propose an auto-weighting deep-learning fuzzy collaborative intelligence approach to address these issues. The experts' authority levels are automatically assigned based on their past performances in the proposed auto-weighting deep-learning fuzzy collaborative intelligence approach. The fuzzy weighted intersection (FWI) is applied to aggregate experts' fuzzy estimates based on the assigned authority levels. Subsequently, a deep neural network (DNN) is constructed to defuzzify the aggregation result. We then apply the auto-weighting deep-learning fuzzy collaborative intelligence approach to a case of estimating the unit cost of a dynamic random access memory (DRAM) product. We show the advantage of the proposed approach over several existing methods using mean absolute percentage error (MAPE).