Reliable and efficient methodologies for optimal feature selection and classification using combinatorial approaches

Abstract

In Biomedical data mining dataset contains huge information, which makes tedious for data processing Feature selection and classification is an important in knowledge learning analyses to make decisions in real world datasets This plays an important role of identifying as well predicting results to categorize the dataset Mainly the implementations are carried out to be supportive for bio medical datasets Feature selection and classification is well supportive for bio medical data processing in high imperative high dimensional dataset Due to various challenges in research the existing implementations are not well sufficient to predict classification accuracy To solve the classification problem we need dedicative approaches using the neural network and classification model Initially we propose a feature selection called the multiple kernel fuzzy c means clustering with rough set theory RS MKFCM algorithm and then followed by classification using Cuckoo search and Genetic Algorithm CGA based Neural Network NN classifier At first the multiple kernel fuzzy c means clustering with rough set theory RS MKFCM is projected to classify the micro array high dimensional dataset to select the important features further to apply CGA NN classifier Selected feat ures from micro array datasets are collected and fed to the neural network for training In neural network we utilize scaled conjugate gradient algorithm for training It provides faster training with excellent test efficiency To improve the classification performance hybridization of cuckoo search and genetic algorithm CGA is utilized with neural network for weight optimization process At last the experimentation is performed on the five different micro array dataset newline