Classification and Feature Selection of Symptomatic and Climatic based Malaria Parasite Counts Using Support Vector Machine

Abstract

Dynamics of Malaria parasite transmission is complex and been widely studied. Research is needed to find a subset of the original features, that will generates a classifier with the highest possible accuracy. Feature selection improves classifier performance; because some machine learning algorithms are known to degrade in performance when faced with many irrelevant/noisy features. In this paper, Support Vector machine (SVM) with One_against_all algorithm is employed to select optimal features for the multiclass symptomatic and climatic malaria parasite-count. Monthly surveys of malarial incidences cases were collected from sampled health centers in Minna Metropolis, Niger State, Nigeria and served as input variables. Linear, Radial Basis and polynomial kernel function were employed but SVM with radial basis kernel function produced better performance result of 85.60% Accuracy, 84.06% Sensitivity and 86.09% Specificity at optimum threshold value of 0.60. SVM selected optimal features to improve prediction performance and reduces time complexity. The experimental results show the robustness and reliability of the proposed model compared to the previous related models.