In many applications we must make a model from the behavior of a man or a group of people. For example, prediction of a gang behavior to find what is their next target. For these kind of examples how can we predict?
You can find a lot of examples for modelling human behaviour in the "Journal of Artificial Societies and Social Simulation" which is an open access journal (http://jasss.soc.surrey.ac.uk/).
Furthermore, there is a lot of material available from the European Social Simulation Association (ESSA) website (http://www.essa.eu.org/). Try to get hold of some of the ESSA conference proceedings. You could search in Google Scholar for "essa+proceedings+simulation+group+dynamics+fight" (http://scholar.google.co.uk/scholar?q=essa+proceedings+simulation+group+dynamics+fight).
Dear Dr. Nouri, the answer very much depends on what abstraction level you want to model humans. We have done research in many aspects, but we always faced an informational or representational complexity which non-linearly grows with the increase of the requested fidelity. The two papers I attached below may explain you why this is the case. Modeling complex social behavior, in particular, perception, cognition, or emotion driven behavior of individuals, groups of individuals, or crowds of individuals, is an emerging research interest. We believe that the results are still in their infancy. However, there are already some important results available such as that is presented in the third paper.
I would suggest you take a look at modeling terrorist behavior and take some of the modelingstrategies and adapt them to Gang related targeting. For example look at:
*Stochastic Modeling of a Terrorist Event via the ASAM System* by Singh, Allanach, Tu, Pattipati, Willett
Adding to Hernando's comment, you can take a look at Ellen Riloff's work. For the topic of crime related narrative tagging, you might want to look at :
*Riloff, E. and Schmelzenbach, M. (1998) "An Empirical Approach to Conceptual Case Frame Acquisition", Proceedings of the Sixth Workshop on Very Large Corpora (WVLC-98).
While other methods of text categorization exist, this is a good place to start.
I think your question should be more precise. in order to obtain valid answer you have to add for what you need this model. In general, every element of the real world may have quasi infinite number of models (see the TOGA meta-theory) In other way every response is OK..
I am just back from the ESSA Summer School for Social Simulation. There the lecturers presented several different approaches of modelling human behaviour. Many of the lectures were recorded and can be downloaded from: http://www.tuhh.de/essa/lectures-material.html (the missing videos might be added in the near future - so make sure you check back some time).