A Framework for Mapping Crime Data on Sociological Hypotheses

Abstract

This thesis proposes a novel framework that helps in getting a better understanding of crimes from a societal perspective. This understanding helps in finding the breakdown of society’s structure and thus can be of great value to decision-makers and reformers for devising crime reduction and crime mitigation strategies. As we know, crime is a serious concern as it affects the economical, emotional, and mental state of others living in the same communities. We all also witnessed that the crime rate has been on a rise all over the world over the years. This increase also makes it essential for data scientists and software engineering to use their skills and background for getting a better understanding and insight into the crime happening around us. The thesis proposes a framework to incorporate already established social criminal theories (SCT) for analyzing the crime data. This novel framework digitally links the SCT being worked on for many years by sociologists to the data record of criminals. The datasets that are used for the proposed methodology include the criminal case dataset, SCT structured form data, and census dataset (containing demographic information of districts). The attributes related to criminal cases are extracted from the criminal record of convicts which are available in multiple forms related to crime. The proposed framework presents a methodology to characterize SCT on a set of attributes. This structured form of SCTs is validated by the subject experts. The discernability of these attributes for each SCT is established by finding the Euclidean distance of each SCT from others and ensuring all the theories are distinctly mapped in the attribute space. The thesis then proposes a novel fuzzy logic-based approach for mapping criminal data on SCT attribute space. The thesis performs several experiments to validate the proposed framework and methodology. First, a number of individual criminal cases are mapped to SCT attribute space to find out the closed association of the case with a specific SCT. The experiments are then extended by mapping several criminal records of convicted criminals on SCT attribute space. This is achieved by first geo-clustering the crime data and then mapping each cluster head to SCT attribute space. The thesis also presents the development of a knowledge base, where reasoning about the number of crime cases with attributes is listed with the help of experts. The report is generated at the cluster level, where the information for the report is extracted from the developed knowledge. The results are validated by presenting them to a panel of domain experts. Moreover, the criminal case studies are also taken from newspapers and map them into the respective social criminal theory. According to the findings, the most applicable SCTs based on the available data are social disorganization theory and social control theory. The other theories are also applicable but to a limited extent. The results demonstrate the effectiveness of the proposed framework. Remedial actions as proposed by sociologists are also suggested to bring changes in society that can help in the reduction of particular types of crimes prevailing in any segment of the society in a locality. This novel work can lead law enforcement agencies and relevant government organizations to cluster crimes being committed in their respective communities to one or more SCTs and then take the help of the rich work that has already been done by sociologists in curbing crimes in communities through the right intervention of policies, incentives, and enforcement. Moreover, the proposed i methodology can be used to identify new social hypotheses that are still hidden from sociologists’ studies by expanding the analysis of digital crime data.