A Computer Based Approach for the Prediction of Multiclass Symptomatic Malaria Infection

Abstract

One of the major public health problem is Malaria infection, accounting for the death of millions of people every year apart from contributing to economic backwardness. The large number of deaths recorded with malaria is as a result of many factors includes: Poor diagnosis, self-medication, shortage of medical experts and insufficient hospital medical laboratories. Therefore, the need for an enhanced malaria expert system is greatly needed. An Artificial Neural Network machine learning technique was used on the set of malaria conditional variables to generate explainable rules. The labeled database was divided into four different levels of severity and classes in Malaria. Out of 14 data that the physician considered as positive, the ANN found that 11 were positive and 3 were negative. Moreover, out of the 11 data that the physician considered negative, the ANN found that 2 were negative and 9 were positive. Therefore, The ANN produces classification accuracy 65.22% accuracy, 57.89% specificity and 100% sensitivity with malaria infection on both the training set and testing set. Further studies will focus on using different machine learning techniques to handle multiclass infection cases.