Crime pattern recognition based on high-performance computing

Eissa, Ahmed, Cheng, Xiaochun ORCID: https://orcid.org/0000-0003-0371-9646 and Petridis, Miltos (2018) Crime pattern recognition based on high-performance computing. In: 2017 International Conference Next Generation Community Policing, 25-27 Oct 2017, Heraklion, Crete, Greece. . [Conference or Workshop Item]

PDF - Final accepted version (with author's formatting) Restricted to Repository staff and depositor only Available under License Creative Commons Attribution 4.0. Download (1MB)

Abstract

This paper reported research results on capturing the semantic similarity of two crimes categories based on the crime description. This study is based on the police data, collected from different sources. The proposed solution can capture the semantic similarity distance between the crimes from the crime database in order to put the selected crime in the right category. We focussed on the crime features and the description of each crime. We will also implement one of Deep Learning algorithms the Recurrent Neural Networks (RNN) algorithm, on top of a labelled crime data that is captured from online police data source. The crime data are labelled using a previous solution presented in a previous work in order to prepare the data to be introduced to RNN algorithm. The proposed solution can capture the semantic similarity distance between the crimes from the crime database in order to put the selected crime in the right category. We focussed on the crime features and the description of each crime. In elaborated experiments, the researchers built tools based on RNN as well as proposed algorithm from previous research, which implements RNN on a top layer and Natural Language Processing based Text Mining on the lower layer. The experiment results so that there is a speed up in RNN training time with an increasing number of threads and CPU cores used. In an elaborated experiments, the researchers built tools based on a proposed algorithm in this research, which implements Natural Language Processing based Text Mining techniques.

Actions (login required)

View Item

Statistics

DownloadsShow export options

Export as CSVXMLJSON

Activity Overview

6 month trend

14Downloads

6 month trend

501Hits

Additional statistics are available via IRStats2.