Nonlinear regression model based on Choquet integral with ε -Measure

When the sub-tests of a composite test are with interaction, the performance of the traditional additive scale method is poor. A nonlinear regression model based on fuzzy integral with the well known fuzzy measures, lambda-measure and the epsiv-measure proposed by our previous study can be applied to improve this situation. In this study, a real data experiment by using a 5-fold cross-validation mean square error (MSE) is conducted. The performances of two nonlinear regression model based on Choquet integral with epsiv-measure and lambda-measure, a ridge regression model, and a multiple linear regression model are compared. Experimental result shows that the performances in order are the nonlinear regression model based on Choquet integral with epsiv-measure, the nonlinear regression model based on Choquet integral with lambda-measure, the ridge regression model, and the multiple linear regression model.