The Research On Ensemble Incremental Learning Classification Algorithm

With the rapid development of the Internet,mobile communications,Internet of Things and other fields,the data generated is exponentially growing,and the form of data is changing from static dataset to dynamic data stream.How to quickly and effectively acquire the implicit knowledge from data stream becomes particularly important.In real life,Non-stationary environment may change the knowledge contained in the data.Traditional batch learning methods encounter many challenges in dealing with dynamic data streams.Therefore,it is an urgent need to design an incremental model to process dynamic data stream in real time.Incremental learning is an important research direction in the field of machine learning.It can learn new knowledge while retaining previous knowledge to a certain extent.The introduction of ensemble learning effectively improves the effect of incremental learning,but in the process of ensemble learning,in order to improve the diversity of the base classifier,the introduction of randomness in the training process of the base classifier will lead to some redundant and poor performance base classifiers in the ensemble classifier,and with the increase of the scale of the ensemble classifier,a large number of models will not only consume a lot of memory.Storage will also take more forecasting time.In this paper,an ensemble model is constructed by multiple base classifiers to incrementally learn the new data.The aim is to establish an efficient and accurate incremental learning model for predicting data stream.This paper mainly completes the following tasks:(1)Introducing the idea of ensemble learning into incremental learning can improve the learning effect.In recent years,most of the research on ensemble incremental learning combines multiple homogeneous classifiers with weighted voting method,which does not solve the problem of stability-plasticity in incremental learning very well.An incremental learning algorithm based on heterogeneous classifier ensemble is proposed.In the stage of training,to make the ensemble model more stable,many base classifiers are trained with new data and then append into heterogeneous ensemble model.Meanwhile,Locality-Sensitive Hashing was used to save the data sketch for the nearest neighbor search of test sample.In order to adapt to the changing data,the newly acquired data will be used to updates the voting weight of the base classifier in the ensemble model.In the prediction stage,for the class label prediction of the test sample,the data similar to the test sample is found from the Local-sensitive Hashing table,this data is used as a bridge to calculate the dynamic weight of the base classifier for the test sample.Determine the class label of the test sample by combined the voting weight and dynamic weight of many base classifiers.Through comparative experiments,it is proved that the proposed algorithm has well stability and generalization ability.(2)With the increase of learning times,the scale and complexity of the traditional ensemble incremental learning algorithm are increasing.To solve this problem,an incremental learning algorithm based on selective learning is proposed.In the training process,the new data are incrementally learned by the base classifier in the ensemble model.If the classification accuracy of the classifier on the latest data set is improved after incremental learning,the incremental learning model will replace the base classifier,or new base classifier will be trained and add to ensemble model.I f the number of classifier reached predetermined threshold,remove the oldest base c lassifier.In the classification process,each classifier is combined with the base classifier by Stacking method to get the ensemble model.Through comparative experiments,it is proved that the incremental algorithm has small scale,high stability and generalization ability.