A selective ensemble method based on K-means method

As the ensemble learning is able to combine multiple weak machine learning methods into a stronger ability, it is widely concerned by the academic community. At the same time, it also has the disadvantages such as large storage space, long training time and repeated training. In order to overcome these drawbacks, we propose an alternative approach based on K-means. The new method is to change the view of the ensemble learning about the integration of all sub algorithms. It is based on the k-means method to select the representative algorithms for algorithms library, so that the redundant or duplicate sub algorithms can be filtered out. In the experimental part, our approach is compared with seven classic methods, including BP neural network, Decision tree, Random forest andfour kinds of integration method. From the comparison of the results, the new algorithm can reduce the complexity of the computation, and also has obvious advantages in the accuracy.