Prediksi Kejadian Malaria dengan Pemanfaatan Data Meteorologi Menggunakan Gated Recurrent Unit

Abstract

Malaria, a disease caused by mosquitoes infected with Plasmodium parasites, continues to be a serious global health issue. Despite efforts to reduce malaria cases, there has been a significant increase, especially during the COVID-19 pandemic. The World Health Organization (WHO) has set a target through the Global Technical Strategy (GTS) to substantially reduce the number of malaria cases, but as of now, this target has not been achieved. To attain this goal and control the spread of this disease, it is crucial to predict the total estimated malaria cases that will occur and take appropriate preventive actions based on these predictions. This research employs inferential statistical approaches to analyze meteorological variables that influence total malaria cases and utilizes a deep learning method known as the Gated Recurrent Unit (GRU) to predict malaria occurrences for the next 12 weeks. The data used includes meteorological parameters such as rainfall, air temperature, and wind speed. The malaria incidence data analyzed originates from Batu Bara District, North Sumatra, collected daily throughout the year 2020. To complete the dataset, this study will also synthesize data for the next three years using Conditional Tabular Generative Adversarial Networks (CT-GAN). The best-performing model achieved a training loss with a Mean Squared Error (MSE) of 0.008 and a validation loss with an MSE of 0.025. The model parameters used encompass a maximum of 500 epochs, 64 hidden neurons, a batch size of 8, a learning rate of 0.001, L2 regularizers at 0.01, and the Adam optimizer