The Systematic Review of Multinomial Naive Bayes (MNB) Based-Model for Detecting and Grading of Malaria Disease

ABSTRACT

Malaria is mostly a disease that is transmitted by the bite of an infected female Anopheles mosquitoes and it affects people mostly in tropical and subtropical countries. Although, previous researchers have tried sourcing for solution to this illness using symptomatic apparatuses, such as Rapid Demonstrative Tests (RDTs), but the study still prove that they are all intrusive and resource-dependent owing to the scalability and environmental features not factored in. This is a big lacuna to the determinant of Malaria disease. The diagnosis of this disease is not readily accessible to everyone in Nigeria, especially where the number of patients are enormous, compared to the number of medical personnel available. In all realities, the earlier researches conducted was based on diagnostic methods which rely heavily on patient-provider interaction and physical sample collection, without environmental factors consideration as well. Therefore, systematic literature review of Multinomial Naïve Bayes model for detecting and grading malaria disease is considered as one of the recent research areas in the world of medical diagnosis applications. The problem of Malaria Disease diagnosis was solved as a multi-class classification solution where there are various levels of the disease. 0(Negative) implies No Malaria, 1(Positive) means there’s Malaria, which was further classified based on severity from Malaria Outcome dataset: Parameter Indicator above and equal to 70% - Complicated and below 70% -Uncomplicated. In this study, a framework for early detection and grading was designed followed by formulating the mathematical model for feature selection of MNB model, an algorithm for detection and grading was developed and implemented, and then, evaluation and comparison with another model (Random Forest) was done to see how MNB model performed.

The study designed the framework by considering the dataset attributes in relation to the environmental parameters. Secondary dataset from clinical records of patients at a public hospital in Abuja were used for training and testing in 80% and 20% ratio. Furthermore, a mathematical formulation was deduced according to the proposed model used, followed by classifier algorithm development, which were implemented through an interface. The result produced a performance accuracy of 97% as opposed to other model used, (RF) which gave rise to inconsistent accuracy (100%) result. This study proves that using MNB model is highly efficient since it produced representative features for the used dataset.

Keywords: Malaria Disease, Grading, Hospitals, Supervising Machine learning, Multinomial Naïve Bayes (MNB)