Improving prediction accuracy of classification model using cascading ensemble classifiers

Abstract

One way to improve the accuracy of predictive modeling is by combining the models. This research tries to study local cascade. It combined one or more base classifier sequentially. In each stage, the probability prediction of the base classifier was inserted to the data. The data then modeled using a decision tree algorithm. This process continued until the data is homogenous. In the original method, the base classifier used was non-ensemble classifier. Our study included bagging, boosting, and random forest as base classifiers. 11 dataset with binary response was used to assess the accuracy of this method. We also compared the accuracy of our method with others that were published between 1996 and 2009. We found that cascading ensemble classifier slightly improve accuracy and performed better for a dataset with numerical predictors.