Proposition of innovative and scalable information system for call detail records analysis and visualisation

Abstract

Call Detail Records (CDRs) are one of the most popular information sources used in criminal investigations. They allow police officers to quickly identify the key actors and relations between them. Of course, the challenge for law enforcement officers is to process and understand the large volume of such data. Typically, the process is long and mostly manual, since the current in-house tools are not well suited for big data analysis and do not use machine learning advances. Therefore, in this paper we investigate how the BigData tools can be used in order to analyse and visualize these kind of data in a scalable and efficient manner. In particular we train and compare various discriminative Machine Learning (ML) models to detect behaviours that are related to criminal activities. In our experiments with the realistic dataset we obtain and report the promising results.