Fuel theft is a concern faced by various countries and oil and gas companies due to its impact on the environment and the safety of communities near oil pipelines. Thus, monitoring and inspecting the location of the oil pipelines through alerts is essential to prevent suspected fuel theft events and mitigate risks. Alerts are triggered by monitoring systems, and patrols are sent to verify the location and confirm the occurrence of illegal tapping. However, various signals can be activated in a short period, and their correct prioritization is essential to identify illegal tapping as quickly as possible. This work aims to use data science and machine learning techniques to perform a predictive model capable of forecasting the probability of an event resulting in an illegal tapping and understanding the factors that influence the occurrence. A Brazilian oil and gas transportation company provided data from a monitoring system supervised from January 2019 to August 2021. We used Four machine learning algorithms: Logistic Regression, Random Forest, XGBoost, and Catboost. The Random Forest obtained the best results in classifying alerts associated (or not) with an illegal tapping., showing accuracy and specificity of 78.6% and 68.3%, respectively. In this problem, specificity means that it is possible to reduce the sending of patrols on the field in 68.3% of cases. As for the external validation, the model also showed a good performance, with an accuracy and specificity of 61%. The high duration of the alert, the high history of illegal tapping, and the occurrence during the night are the ones that most influence illegal tapping occurrence.