An Application of the Naïve Bayes Algorithm as a Tool for Predicting Cases of Poaching in Wildlife Management

Abstract

Poaching presents a serious threat to both wildlife management and tourism sustainability, making proactive, data driven interventions necessary. This study explores the application of predictive data mining in crime management within Zimbabwe’s wildlife sector, focusing on the use of the Naïve Bayes algorithm to predict poaching occurrences in protected areas. Secondary data from ranger patrol logs, incident reports, weather observations, and spatial datasets were collected from selected wildlife management areas between 2015–2023. The data was cleaned, integrated, and underwent feature engineering before it was modelled with the Naïve Bayes classifier to identify patterns of poaching risk. After evaluation, the model achieved an accuracy of 87%, precision of 0.84, recall of 0.82 and a ROC (AUC) curve of 0.89 showing strong predictive capabilities. Key predictors included weather conditions, patrol intensity and recent poaching history. The model’s false positives (predicting poaching where none occurs) may result in extra patrols, while false negatives (failing to predict actual poaching) put wildlife at greater risk, showing the importance of balancing prediction sensitivity and resource allocation. These findings show that predictive data mining can promote crime prevention, enhance resource allocation, and reinforce wildlife protection in Zimbabwe. The study recommends including predictive analytics into conservation planning to protect the country’s wildlife resources.