A new ensemble model based on linear mapping, nonlinear mapping, and probability theory for classification problems

Currently, various perspectives of neural networks are proposed for solving classification problems. Some of them are based on two types of mapping functions, namely, linear and nonlinear, for mapping an input space into a feature space. In addition, some neural networks are proposed based on probability theory. Since some models are appropriated for some kinds of data, depending on a distribution of the data, some data are appropriated for linear mapping, some is for nonlinear mapping, and some is for probabilistic models. Due to the fact that the data distribution in classification problems are various, we propose the new ensemble model based on linear mapping, nonlinear mapping, and probability theory for classification problems. According to the experimental results, they have shown that our proposed model can improve the accuracy of classification on the testing data sets.