Diversity Matrix Based Performance Improvement for Ensemble Learning Approach
This chapter provides a novel way of solving the performance deficit in the electroencephalography (EEG) based bio-signal classification with a satisfactory performance in its first adaptation. It explains related work on the EEG signal classification and also provides the necessary theoretical understanding for our study. The chapter explains the proposed diversity matrix based pruning technique with multiple examples. A highly popular independent ensemble classifier method is bagging. Cosine similarity measures the similarity between two non-zero vectors. Kullback-Leibler divergence, which is also called relative entropy, is related to information divergence and information for discrimination. The prime focus of the proposed method is to improve the already obtained classification accuracy from an ensemble by majority voting. Mix-bagging is an independent ensemble classification method. The concept behind mix-bagging is to use multiple learners instead of a single type of classifier.