A Choquet integral regression model with a new fuzzy measure based on multiple mutual-information

The well known fuzzy measures, lambda-measure has no information with the dependent variable. Owing to above problem, the epsilon-measure based on multiple entropy is proposed by our previous study. In this paper, an improved fuzzy measure based on multiple mutual-information, called M-measure, is proposed. For evaluating the Choquet integral regression models with different fuzzy, measures, a real data experiment by using a 5-fold cross validation mean square error (MSE) is conducted. The performances of the Choquet integral regression models based on M-measure, epsilon-measure and lambda-measure, respectively, a ridge regression model, and the traditional multiple linear regression model are compared. Experimental result shows that Choquet integral regression model based on the new measure, M-measure, has the best performance.