While studies typically investigate the socio-economic factors of perpetrators to comprehend terrorism motivations, there was less emphasis placed on factors related to terrorist attack locations. Addressing this knowledge gap, this study conducts a multivariate analysis to determine attributes that are more associated with terrorist attacked locations than others. To tackle the challenge of identifying pertinent attributes, the methodology merges a global terrorism database with relevant socio-economic attributes from the literature. The workflow then trains 11 machine learning models on the combined dataset. Among the 75 attributes assessed, 10 improved the predictability of targeted locations, with population and public transportation infrastructure being key factors. After optimizing hyperparameters, a multi-layer perceptron—a type of artificial neural network—exhibited superior predictive performance, achieving an AUC score of 89.3%, classification accuracy of 88.1%, and a harmonically balanced precision and recall score of 87.3%. In contrast, support vector machines demonstrated the poorest performance. The study also revealed that race, age, gender, marital status, income level, and home values did not improve predictive performance. The machine learning workflow developed can aid policymakers in quantifying risks and making objective decisions regarding resource allocation to safeguard public health.