The Prediction of Leptospirosis Outbreaks in Europe with Artificial Intelligence

Leptospirosis is a reemerging zoonotic disease caused by bacteria of the Leptospira genus, posing a significant public health threat in both developed and developing countries, it affects approximately 1.03 million people and results in an estimated 58,900 deaths annually. In Europe, confirmed cases saw a notable increase in 2018, exceeding 1,000 cases, following a period of stability and a resurgence in 2014. However, recent data shows a decrease to 765 cases in 2022, likely influenced by the habits developed during the COVID 19 pandemic. This thesis conducts a spatio-temporal analysis of ECDC Leptospirosis Europe patient data, utilizing NUTS3 for spatial resolution and monthly temporal resolution to identify infection regions and periods. The study incorporates 22 climate features derived from temperature and precipitation, including mean monthly temperature, minimum and maximum temperature, mean weekly temperatures, monthly temperature range, monthly total precipitation, minimum and maximum precipitation, total weekly precipitations, and precipitation range, as well as mean temperatures of the driest and wettest weeks, and total precipitation of the coldest and hottest weeks. Additionally, the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Water Index (NDWI) were considered as environmental drivers. Socio-economic factors, including population density, employment, and Gross Domestic Product (GDP), were also examined to assess their impact. Balanced datasets were created from the infection and control records, and four XGBoost classifiers were developed to predict disease presence. A soft voting ensemble technique was implemented using the four developed models, achieving an accuracy of 70-72% across the four test sets, highlighting its superior predictive capability. The modeling analyses enhanced understanding of environmental features linked to human cases of Leptospirosis in Europe, contributing to the development of integrated surveillance systems. Further research is needed to establish causal impacts.