Hill Climbing Based Feature Selection and Self Organizing Deep Auto-Encoder (Sodae) Classifier for Snps Dataset

Abstract

Single Nucleotide Polymorphisms (SNPs), are known as "snips" in the genetics community, are the most common kind of genetic variation in humans.A single nucleotide might vary in each SNP, which is a fundamental component of DNA. Significant SNPs in many human illnesses have been found using a variety of Machine Learning methods.Issues in the area of SNP identification include the dimensionality curse and the possibility of uneven numbers of healthy and patient samples. Thus the feature selection and classification has been efficient solution to solve these challenges in the SNP dataset. A novel method is presented in this research to feature selection, and classification of two different SNP datasets. The nominal SNP data are first transformed the mean encoding to numerical data.For feature selection, the Hill Climbing (HC) algorithm was then devised, which eliminates characteristics that are unnecessary and redundant.In order for the classification to create its structure based on input data, a Self-Organizing Deep Auto-Encoder (SODAE) is used. To assess a classifier's performance, Breast cancer and thyroid cancer SNP datasets have been gathered from the Gene Expression Omnibus (GEO). Precision, recall, F-measure, and accuracy for classification techniques are used to evaluate the effectiveness of the findings. When compared to techniques like k Nearest Neighbor (kNN), support vector machines (SVM), and support vector machines recursive feature elimination (SVM-RFE), the findings show that it has been highly successful.