DISEASES SPREAD PREDICTION IN TROPICAL AREAS BY MACHINE LEARNING METHODS ENSEMBLING AND SPATIAL ANALYSIS TECHNIQUES
Keywords: disease spread, machine learning, ensembling, LSTM
Abstract. Infection with tropical parasitic diseases has a great economic and social impact and is currently one of the most pressing health problem. These diseases, according to WHO, have a huge impact on the health of more than 40 million people worldwide and are the second leading cause of immunodeficiency. Developing countries may be providers of statistical data, but need help with forecasting and preventing epidemics. The number of infections is influenced by many factors - climatic, demographic, vegetation cover, land use, geomorphology. The purpose of the research is to investigate the space-time patterns, the relationship between diseases and environmental factors, assess the degree of influence of each of the factors, compare the quality of forecasting of individual techniques of geo-information analysis and machine learning and the way they are ensembled. Also we attempt to create a generalized mathematical model for predicting several types of diseases. The following resources were used as a data source: International Society for Infectious Diseases, Landsat, Sentinel. The paper concludes with the summary table containing the importance of individual climatic, social and spatial aspects affecting the incidence. The most effective predictions were given by a mathematical model based on a combination of spatial analysis techniques (MGWR) and neural networks based on the LSTM architecture.