Background: India has a rising rate of malaria as well as a high mortality rate despite awareness and efforts being focused on the issue. Some regions are profoundly affected than others, such as in Odisha, where the prevalence of malaria is nearly a third of the whole country. This study investigated the influence of climate factors on the incidence of malaria in the Sundargarh district in the state of Odisha, India.
Methods: Block-wise observed station rainfall data was sourced from the Special Relief Commissioners' (SRC) web portal. Gridded surface maximum temperature and relative humidity data were accessed from the European Center for Medium-range Weather Forecast (ECMWF) reanalysis data archive. Malaria incident data were collected from the Directorate of Public Health, Government of Odisha. WEKA machine learning tool with two classifier techniques, Multi-Layer Perceptron (MLP) and J48 with 10-fold cross-validation, percentile split (66%), and supplied test options, were used for the Malaria prediction. A comparative analysis was carried out on both techniques to ascertain the superior model amongst the two, concerning the prediction accuracy of malaria in the context of a varying climate. Classifier accuracy, Root Mean Square Error (RMSE), Kappa, and ROC scores were the indicators used for the analysis.
Results: The results suggested that J48 had exhibited a better skill to MLP and illustrated less error with a positive kappa. In particular, the 10-fold cross-validation method had better performance over the percentile Spilt (66%) and supplied test options. J48 demonstrated less error (RMSE = 0.6), better kappa = 0.63, and higher accuracy = 0.71), suggesting it as most suitable model. Further, seasonal temperature and humidity variation had shown a better association with malaria incidents in comparison to rainfall.
Conclusion: The performance of the machine learning methods for Sundargarh was particularly better during the monsoon and post-monsoon when the events are at the peak. The results were encouraging for the utilization of climate forecast for prediction of malaria incidences. It is thus recommended that the J48 classifier machine learning technique could be adopted for the development of malaria early warning system.