Developing a crime analysis simulator using machine learning methods

Crime analysis plays important role in prevention of crimes, investigate the relations among crimes, detection of criminals, helping police investigations and revealing the events quickly. In addition, it is also important to perform crime analysis for efficient use of human and technical resources. In the thesis, crime analysis is performed by using a forecasting model based on Bayesian network. As a result, decision-making system based on GIS is proposed. So, the synthetic crime data set, is used to find the crime trends and potential criminal among crime incident-level. Generally five contributions are provided in this thesis: First of this contrubution, , a parametric model is proposed to generate the incident-level crime datasets involving crimes, GIS, criminals and criminals' suspicious acquaintances where the parameters are used for fine tuned adaptation of the model. The datasets are generated by modeling the population in real-life. The model can also produce random geographical maps and distribute population, crimes and criminals to the geography based on real-life approximations .Second, the relationship between the crime events and the relationship between the crime is determined in the determination of the relationship. Here, it was determined that the criminals were related to other crimes and the inclinations of the types of crimes were found. Third, a clustering based method is proposed to infer the closely related incidents using hybrid similarity metrics and it is demonstrated a decision-making process using hierarchical clustering algorithms in finding important patterns between criminals' social network and crimes in trend. Fourth, this thesis demonstrate the results of some queries related to crime analysis over the GIS dataset and discover committing criminal on given synthetic dataset. Finally, it is used to detect the relationship between crime groups and reach the leaders of the crime organizations with through criminals and criminal's acquaintances.