Bisecting data partitioning methods for Min-Max Modular Support Vector Machine

Min-Max Modular Support Vector Machines (M3-SVM) is a well-known ensemble learning method. One of key problems for M3-SVM is to find a quick and effective method for data partition. This paper presents a new data partitioning method-BEK, which is based on the Bisecting K-means clustering with equalization function. BEK generally can get global optimal solution with low time complexity, and more importantly, it can obtain the relatively balanced partitions, which are very important for M3-SVM to deal with huge data. Experimental results on real-world data sets show that this bisecting partitioning method can effectively improve the classification performance of M3-SVM without increasing its time cost.