Bininng And Hybrid Feature Selection Based Big Data Classification

Main Article Content

S. Saravanabavanandam , Dr. S. Duraisamy

Abstract

The improved information technology resources, genomics, health records ear creates opportunity for leveraging these developments. A learning health system is created using these developments for delivering informative clinical evidence. An information and data set which is highly complex and large is termed as big data. It is highly complicate to process this big data using conventional database management tools.To overcome those issues in recent work first introduces the pre-processing step using min-max normalization.And then synthetic minority oversampling is used to balancing the data set by generating synthetic data. And Features selection is computed based on levy flight grey wolf optimization additionally introduces hybrid model using CNN with SVM for big data classification. However Data set taken for this work may have noisy data values and it may affect thebig data classification performance and it does not focused in existing work. Performing classification task is based on the single classifiers which not improve the accuracy of the classifier. To avoid those issues this work first introducesthe binning for smoothingand removing unwantedpixels. And then introduces the pre-processing step using min-max normalization. It will normalize the input data into same scale. And then synthetic minority oversampling is used to balancing the data set by generating synthetic data.And then feature selection will be performed based on hybrid Chicken Swarm Optimization andWhale Optimization algorithm. Finally classification doneby usingensemble CNN-SVM. In which Ensembling of the classifier will be done by majority voting functionfor the outputs.Experimental results demonstrates proposed model’s effectiveness using Covtype, ECBDL14-S and Poker database interms of precision, recall, error rate and accuracy metrics by comparing with existing  HMM, FKNN, WCNN, WCNN-SMT and CNN-SVM  using MATLAB.  

Article Details

Section
Articles