Multinomial Logistic Regression with Fuzzy Parameters

Two new developed multinomial logistic regression approach is proposed by incorporating the concepts of fuzzy sets. The first is formulated using a goal programming approach, while the second is formulated as a multi objective programming model. These two models are based on the assumption that the parameters are fuzzy. A simulation study is used to evaluate the suggested models comparing to the classical approach. Data are generated from different multinomial logistic models The design of the simulation study considers 40 different combinations of three factors. For each combination, a comparison between the performance of the proposed approach and ML approach is presented.