Combining nearest neighborhood classifiers using genetic programming

In this paper, GP based intelligent scheme has been used to develop an Optimal Composite Classifier (OCC) from individual nearest neighbor (NN) classifiers. In the combining scheme, first, the predicted information is extracted from the component classifiers. Then, GP is used to develop OCC having better performance than individual NN classifiers. The experimental results demonstrate that the combined decision space of OCC is more effective. Further, we observed that heterogeneous combination of classifiers has more promising results than their homogenous one. Another side advantage of our GP based intelligent combination scheme is that it automatically incorporates the issues of optimal model selection of NN classifiers to achieve a higher performance prediction model.