Cluster analysis of Delhi crimes using different distance metrics

Crime greatly affects the well being of human population. Mitigating and preventing crime are the issues of primary importance. Analysis of the bygone datasets, could reveal important insights, which if properly used can prevent crimes from happening. In this research paper we have considered a subset of the crime dataset in Delhi, focusing on the data of missing children in Delhi for the year 2016, which has possible links to kidnapping. In this data set we have the last seen location of the children before they were reported missing. We have plotted the Geo-location of each locality and used this data for Geo-spatial clustering in k-means using different distance metrics. The performance of each of the distance matrix is compared and the results are used to evaluate the best metric for location based clustering over a small city like Delhi. This work can be taken by the law and security agencies to increase protection in the areas of higher children missing rate. The distance metrics that are used in this project are Haversine and Euclidean.