Incorporating EBO-HSIC with SVM for Gene Selection Associated with Cervical Cancer Classification.

Abstract

Microarray technology is utilized by the biologists, in order to compute the expression levels of thousands of genes. Cervical cancer classification utilizing gene expression data depends upon conventional supervised learning methods, wherein only labeled data could be used for learning. The previous methodologies had problem with appropriate feature selection as well as accurateness of classification outcomes. So, the entire performance of the cancer classification is decreased meaningfully. With the aim of overcoming the aforesaid problems, Enhanced Bat Optimization Algorithm with Hilbert-Schmidt Independence Criterion (EBO-HSIC) and Support Vector Machine (SVM) algorithm is presented in this research for identifying the specific genes from the gene expression dataset that belongs to cancer microarray. This proposed system contains phases of instance normalization, module detection, gene selection and classification. By Fuzzy C Means (FCM) algorithm, the normalization is performed for eliminating the inappropriate features from the gene dataset. Meanwhile, for effective feature selection, the EBO algorithm is used for producing more appropriate features via improved objective function values. For determining a subset of the most informative genes utilizing a rapid as well as scalable bat algorithm, this proposed method focuses on measuring the dependence amid Differentially Expressed Genes (DEGs) as well as the gene significance. The algorithm is dependent upon the HSIC and was partially enthused by EBO. With the help of SVM classifier, these gene features are categorized very precisely. Experimentation outcomes demonstrate that the presented EBO with SVM algorithm confirms a clear-cut classification performance for the given gene expression datasets. Hence the result provides higher performance by launching EBO with SVM algorithm to obtain greater accuracy, recall, precision, f-measure and less time complexity more willingly than the previous techniques.