Predictive Analytics for Safer Smart Cities

The threat of incendiary and often, catstrophical terrorist attacks is a major challenge for the urban administrators. The urban landscape is changing at a fast pace with the emphasis moving toward "smart cities". Terrorists, for obvious reasons, prefer attacking cities compared to rural areas. Smart cities are expected to absorb larger populations of inhabitants in smaller area implying the damage inflicted by these attacks would be maximum unless some preventive mechanisms exist in the smart city ecosystem. There exist very few methodologies for attack forecasting due to lack of real-time data (confidentiality and reluctance of law enforement in sharing data). In this paper, we propose a way to predict future attacks, weapons used and likely targets using a class of powerful machine learning algorithms known as ensemble learning. The features used to train the model are location, attack type, weapon type and target type.