Markov Random Field For Threat Detection

A research project collaborated with Professor Michael Baron from American University.

Statistical Research Assistant for the “Algorithms for Threat Detection” Project in Jan. 2019 - Dec. 2019

• Used Markov Random Field to build logistic regression models from 911 call data during 2015 Baltimore riots
• Applied pseudo-likelihood estimation to maximize essential parameters to detect change points
• Predicted possibility of large-scale violence through detection of change-points in Markov Chains
• Conducted tests such as VIF, Stepwise Variables Selection, Linear and Quadratic Discriminant analysis, KNN, and Random Forest to cross-validate and further optimize the model