CLASSIFICATION OF MALARIA PARASITE (PLASMODIUM FALCIPARUM) SEVERITY LEVEL USING MULTI-LAYERED PERCEPTRON AND RADIAL BASIS FUNCTION

Abstract

The purpose of this research is to classify malaria parasite (Plasmodium falciparum) based on its severity level using artificial neural networks (ANNs). This is consequent to the fact that malaria parasitemia remains a lethal force in Africa, killing a child in every two minutes. This study engaged multi-layered perceptron (MLP) and radial basis function (RBF) as ANNs employed for the classification procedure. Available literature affirms the use of various ANNs have been deployed in various medical classification procedures. Nevertheless, it is apparent that studies have not focused on classifying this disease based on its severity level. MLP is a type of ANN designed to have several hidden layers, and has a high learning potential with a proven record of high capability in generalizing tasks. RBF is able to train in maximum likelihood framework by maximizing the probability, thereby minimizing the error. The data engaged in this study were obtained from Federal Teaching Hospital, Ido-Ekiti, Nigeria. 499 observations of malaria parasite were trained, tested and classified using MLP and RBF. The result obtained shows that both MLP and RBF are efficient in classifying malaria severity level. However, MLP has a better classification with 100% accuracy, precision, recall and f1-measure as compared with RBF which has 98.67% accuracy, 99.41 precision, 97.62% recall and 98.47% f1-measure. Conclusively, an accurate recognition and classification of malaria severity level using MLP will assist in combating the parasite early enough before resulting in termination of life.