Methodologies for homicide analysis and prediction:

In this seminar 3 different methodologies for signal processing based on homicide data were presented. In Zero-Inflated Embeddings, embeddings of the different sectors and context vectors are built to help in the prediction process. On the other hand, Constrained Graphs take ideas from two different methodologies, Kernel Warping and GLoG. These graphs include data from related crimes such as theft or personal injury. Finally, a descriptive analysis using the temporal evolution of these graphs is presented.