Machine Learning Approach to Identify the Incidence of Human Deaths Due to Human-Elephant Conflicts

Human-Elephant Conflict (HEC) is one of the predominant issues Sri Lanka is coping with, where the North Central Province (NCP) has recorded the highest human deaths due to HEC. However, nowadays, most travelers are cautious of attacks by wild elephants while traveling through the streets and main roads in the NCP, as elephant threats are ubiquitous. This study aims to identify the environmental and human physiological factors contributing to incidents of human deaths caused by elephant attacks, and to predict the incidents of human deaths using some machine learning techniques, including Classification Trees (CT), Support Vector Machines (SVM), and Probabilistic Neural Networks (PNN), in the NCP. The Department of Wildlife Conservation (DWC) and the Department of Meteorology were contacted to collect data. On the other hand, the agricultural calendar was utilized to obtain information about the paddy harvest season. After identifying the key influential factors for human death incidences, the Chi-square test and Fisher’s Exact test were employed to check the association between the response variable, incidents of human deaths and influential factors. Human gender, age group, rainfall season, paddy harvest season, and time period for encountering elephants were identified as significant factors for the response variable, incidents of human death. Overall Classification Accuracy (OCA) and Balanced Accuracy (BA) were used to measure the performance of each model, in which 80% of the data were randomly selected for the model-building process, and the remaining 20% were used to check model accuracy. Even though all the models demonstrated more than 85% OCA and BA, PNN exhibited the highest OCA of 88.57% and a BA of 88.39%, making it the better model to predict the incidents of human death due to elephant attacks. The final model and findings of this study can be used to ensure the safety of human lives in different incidents, as it reduces uncertainty of life.