Ensemble Learning Study And Its Application

Ensemble Learning is machine learning function. It is a research focus of the current. Ensemble learning has a very good application in many fields. However, the Ensemble Learning technologies have not yet matured, In the course of the study there are still many unresolved issues and the application also falls far short of the level of people's expectations.This paper carried out in-depth study of Ensemble Learning. In this process, On the one hand, analysis of the strengths and weaknesses of existing methods, On the other hand, studied the effects of various factors that affect the integration. This part is divided into pave the way to further improve the Ensemble Learning. Based on existing algorithms, the paper proposed new Ensemble Learning methods:Some of the individual classifier has no effect on the integration of effects or individual classifiers to reduce integration efforts, in response to these circumstances, this paper proposed "Classifiers Selection Based on Information Gain" This method first need to construct classifier space, then calculate the individual classifiers IG, finally integration of the remaining classifiers. Experimental results show that effects of integration has been greatly improved results. Not all features have a positive effect on the integration in a dataset, for this, this paper proposed "Feature Selection Based on Improved Genetic Algorithm" The approach will combine machine learning and genetic algorithms to achieve their goals of feature selection in effective, And then use the integrated learning in the processed data sets. Results show this method not only increased the effects of integration in greatly, but also shorten training time. At last, the paper research on the application of an integrated learning, Firstly, Studied the intrusion detection data set (KDD CUP 99,UCI). Secondly, proper handling of data sets, and then make ensemble learning applied to this data set. Thirdly, Comparison the results of the classical ensemble learning methods and the proposed integrated learning method for intrusion detection data set.The paper validated this new method through many experiments and application examples, the results show that the new method is effective, and feasible.