A Novel Map Reduced Based Parallel Feature Selection and Extreme Learning for Micro Array Cancer Data Classification

Microarray-based gene expression profiling is an emerging method to predict, classify, diagnose and to treat cancer efficiently. The characteristics of this cancer disease may change frequently which creates large volume of data. In this paper we propose a Novel Map reduced based parallel feature selection and extreme learning for micro array cancer data classification. Initially the gene expression data sets are pre-processed by attribute-wise normalization and setting thresholds on the original data. The second phase uses a wrapper model that uses Adaptive Whale Optimization Algorithm (AWOA) with Nelder–Mead algorithm (NMA) to accomplish the feature (gene) subset selection. Wrapper models are used to describe the selection process of feature sets as a search issue. Here, various combinations are formulated, estimated and compared with other combinations. At last, to demonstrate the effectiveness of the selected genes using the proposed feature selection method, a Regularized Extreme Learning Machine (RELM) classifier is used to classify the gene expression data subsets chosen by AWOA algorithm.