Behavioural analysis of crime against women using a graph based clustering approach

Abstract

Crime against women is increasing at an alarming rate in almost all parts of India. Women in the Indian society have been victims of humiliation, torture and exploitation. It has even existed in the past but only in the recent years the issues have been brought to the open for concern. According to the latest data released by the National Crime Records Bureau (NCRB), crime against women have increased more than doubled over the past ten years. While a number of analyses have been done in the field of crime pattern detection, none have done an extensive study on the crime against women in India. The present paper describes a behavioural analysis of crime against women in India from the year 2001 to 2014. The study evaluates the efficacy of Infomap clustering algorithm for detecting communities of states and union territories in India based on crimes. As it is a graph based clustering approach, all the states of India along with the union territories have been considered as nodes of the graph and similarity among the nodes have been measured based on different types of crimes. Each community is a group of states and / or union territories which are similar based on crime trends. Initially, the method finds the communities based on current year crime data, subsequently at the end of a year when new crime data for the next year is available, the graph is modified and new communities are formed. The process is repeated year wise that helps to predict how crime against women has significantly increased in various states of India over the past years. It also helps in rapid visualisation and identification of states which are densely affected with crimes. This approach proves to be quite effective and can also be used for analysing the global crime scenario.