Imbalanced Data Classification Algorithm Based On Hybrid Model

With the rapid growth of the generated data in modern society, it becomes more and moreimportant to extract useful information. Traditional algorithms have a lot of shortcomings whendealing with imbalanced data in classiifcation, and the recognition rate of the rare class usuallycan’t meet people’s need.In this thesis, we introduce a CBP-SVM algorithm that based on hybrid model to improvethe recognition rate of the rare class. CBP-SVM algorithm combines with the advantages ofseveral learning algorithms to deal with the problem of imbalanced data classiifcation: Particle ofSwarm Optimization (PSO) algorithm is used to optimize the RBF kernel parameters of SVM，and the optimized SVM is used as weak classiifer for AdaBoost to get Boost-PSOSVMalgorithm, then combining Boost-PSOSVM algorithm with Cascade Model to get a hybridclassiifcation model to solve the problem of imbalanced data classiifcation. In hybridclassiifcation model，Boost-PSOSVM algorithm makes some modiifcations to the AdaBoostalgorithm then combining with the optimized SVM to improve the classiifcation performance,Cascade Model makes the entire dataset tend to balance by gradually excluding the major classsamples, and Boost-PSOSVM algorithm can focus more on the samples that is hard to classiyf.The experimental results show that CBP-SVM algorithm signiifcantly improves the overallclassiifcation accuracy and the recognition rate of the rare class.