A Comparative Study of Different Classification Techniques on SNP Data

Abstract

Genetic factor is the most important part for consideration, often leads to the analysis of Single-nucleotide Polymorphism (SNP) which actually causes the trait. SNPs arethe large fields of research that have been widely popular in today’s modern world.Now-a-days, different classification methods become widely popular in the field ofgenome wide association study (GWAS) considering significance of SNPs. But thereis no serious study in literature for comparing of different classification techniques. In this paper, it is compared four classical classification techniques with a machinelearning technique for predicting binary trait given the genotypic information. For this purpose, considered simulated data with the help of R packages. The data sets arepartitioned to train and test data consisting 70% and 30% respectively. Then differentclassification techniques are performed, namely logistic regression, linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), naive bayes (NB) and Supportvector machine (SVM). The models are evaluated by the performance measure accuracy, sensitivity, specificity for training data as well as test data. Results suggestthat SVM performs slightly better than other techniques.