Synthetic Minority Oversampling And Levy Flight Grey Wolf Optimization For Big Data Classification

Abstract

Primary disease detection, patient care and community services are greatly benefited through precise medical data exploration which in turn is greatly attained due to tremendous big data evolution in biomedical and healthcare communities. Thereby data volume, velocity, and variation also increase promptly. Henceforth, massive data storing, processing and visualising through classic approaches is a challenging issue.  For mitigating these issues, improved frame work for big data classification where feature selection is performed via adaptive cuckoo search besides classification by weighted convolutional neural network. Conversely, class imbalance problem is yet another challenge for dataset considered in this research which is not concentrated in prevailing research, in addition classifier generalisation ability are also affected. Also, local optimal solution is yet another difficulty faced due to cuckoo search. Synthetic Minority Oversampling Technique (SMOTE) is greatly utilized for mitigating class imbalance problem. The dataset are balanced where numbers of instances are   increased in   minority class along with suggested model error rate reduction. Levy flight grey wolf optimization is greatly involved in feature selection for computation time reduction and thereby classification accuracy is improved. Lastly big data classification is achieved through weighted convolutional neural network. The suggested model is validated through experimental outcomes where effectiveness is demonstrated pertaining to precision, recall and accuracy for Covtype, ECBDL14-S and Poker database.