Analysing Machine Learning Based Models for Predicting Malarial Fever Prior to Clinical Trial

Abstract

Malaria is a very dangerous illness brought on by unicellular protozoan parasites of the species Plasmodium. This disease is endemic in many parts of the world. Confirming the presence of parasites early on in all cases of malaria permits the delivery of species-specific antimalarial medication, which reduces the death rate and points to other illnesses in situations where the diagnosis is negative. Nevertheless, light microscopy of thin and thick peripheral blood (PB) films stained with May-Grünwald–Giemsa (MGG) is still the gold standard. Since this is a labor-intensive process that relies on a pathologist's expertise, medical professionals in regions of the world where malaria is not widespread may have difficulties diagnosing cases of the disease. To predict malaria fever, this research used a total of thirteen different machine-learning models. These models included the Gaussian NB, Logistic Regression, XGB, Bagging Classifier, Random Forest Classifier, Extra Trees Classifier, Gradient Boosting Classifier, Hist Gradient Boosting Classifier, LGBM Classifier, Decision Tree Classifier, Ada Boost Classifier, SGD Classifier, and K neighbors Classifier. To conduct this study, we classified cases of malaria using data obtained from chest X-ray images. A dataset was created using 1079 patient records, and 23 attributes were used. These attributes were gathered from the Kaggle repository. Out of these 23 attributes, 80% of the data were used to train the model, and the remaining 20% were used to assess the validation accuracy. It has been shown that the Gaussian NB models were the most accurate, with a 97.66% accuracy rate.