Employing machine learning and internet of things for malaria outbreak prediction in Rwanda

Abstract

Malaria is a threatening disease which is caused by a bite of female mosquitoes called anopheles and when it is not discovered at its earlier stage, it can put the life of many people at risk and even reduce the workforce of the country. However, its rate of transmission can be decreased if the information regarding the development of these mosquitoes are made available in due time. However, there is a lack of real time information about Malaria spreading to help the Ministry of Health to know the development of malaria mosquitoes relatively to environmental conditions and take the required measures for fighting against the spread of this disease by providing early warning to decision makers, hospitals and health institutions to purchase the medicine on time and reminding the citizens to use mosquito nets accordingly. The current study mainly aims to apply machine learning and Internet of Things technologies to help the Ministry of Health (MoH) to have access on the development of malaria mosquitoes and provide early warning information across citizens, hospitals, health institutions and individuals to be prepared accordingly. For modelling the dependency of malaria transmission, we have tested different machine learning classification algorithms for optimizing the prediction accuracy. The data used include the environmental climate and malaria data recorded by METEO Rwanda and Ministry of Health respectively in the period of 8 years (2012-2019) from Bugesera and Huye districts the most malaria endemic district in Rwanda. The results show that the Artificial Neural Network algorithm could perform better than other algorithms tested with 93.9% and 88.2% of training and testing accuracy respectively in Bugesera district, and 88.9% and 62.5% for training and testing accuracies respectively in Huye district. Secondly, an IoT based system was prototyped to interact with the predictive model and view the results of prediction in the future on field sensors data via Smartphone, tablet or PC.