Predicting Criminal Moments with Hidden Markov Models

We aim to construct a probabilistic model which explains a hypothetical criminal's behavior based on that criminal's actions. We will be given the criminal's motions and the crimes they have committed, and from this, we will try to find a probabilistic model that explains their behavior. We are planning to construct a Hidden Markov Model, which is a probabilistic system describing the transitions between the criminal's actions.

Treating the criminal's decisions about what locations to go to, and which crimes to commit, as the result of a Markov process, we can use statistical methods to reconstruct a model that predicts their behavior. For this project will use a simulated criminal whose actions are generated by an agent-based model (resulting from a previous GSoC project). We should be able to create a model that is nearly equivalent to the original program. With the reconstructed model, the program should be able to make predictions about the criminal's future actions.