Using parallel partitioning strategy to create diversity for ensemble learning

Divide-and-conquer principle is a fashionable strategy to handle large-scale classification problems. However, many works have revealed that generalization ability is decreased by partitioning training set in most cases, because partitioning training set can lead to losing classification information. Aiming to handle this problem, an ensemble learning algorithm was proposed. It used many sets of parallel hyperplanes to partition training set on which each base classifier was trained by the SVM modular network algorithm and all these base classifiers were combined by majority voting strategy when testing. The experimental results on 4 classification problems illustrate that ensemble learning can effectively reduce the descent of generalization ability for the reason of increasing classifier´s diversity.