Crime Analytics: Analysis of Crimes

Through Newspaper ArticlesIndika Perera

Isuru Jayaweera

Chamath Sajeewa

Sampath Liyanage

Tharindu Wijewardane

Department of Computer Science and Engineering

Faculty of Engineering

University of Moratuwa

Sri Lanka

Adeesha Wijayasiri

Department of Computer and Information Science and Engineering

University of Florida

United States

Introduction

Why Crime Analysis?• To identify general and specific crime trends, patterns and

series in an ongoing, timely manner.

• To utilize limited law enforcement resources.

• To be proactive in detecting and preventing crimes.

• To meet the law enforcement needs of changing society.

A major challenge faced by most of the law enforcement and intelligence

organizations is efficiently and accurately analyzing the growing volumes of crime

related data. The vast geographical diversity and the complexity of crime patterns

have made the analyzing and recording of crime data more difficult. Data mining is a

powerful tool that can be used effectively for analyzing large databases and deriving

important analytical results.

Crime Analysis Approaches Using in

SriLanka

• Police department uses manual crime recording and

analysis system.

• There is no free and open accessible crime analysis

system in Sri Lanka.

Grave Crime Records

Minor Crime Records

Usefulness of Crime Analysis System

● Police Department can use the system when they create

security plans.

● Police department can evaluate their existing plans.

● Investors can use the system when they want to find suitable

areas for investments.

● Tourists and tourist agents can use the system when they are

planning their tours.

Related Work

Crime Data Mining Techniques

• Entity extraction

• Association rule mining

• Deviation detection

• Classification

Existing systems that use data mining techniques for

crime investigation

• Regional crime analysis program

• Link analysis concepts

• Data mining framework for crime pattern identification

• Narcotics network in Tucson police department

• Clustering techniques

• Sequential pattern mining

• String comparison

Crime Analysis Steps

• Hotspot Detection

• Crime Pattern Visualization

• Nearest Police Station Detection

A framework has been proposed which includes relationships between the crime data

mining techniques and crime type characteristics. Framework has been developed by

using Tucson Police Department crime classification database. Using this framework,

investigator can determine the most suitable data mining technique for his/her task. As

given by the proposed framework, investigators can use neural network techniques in

crime entity extraction/ prediction, clustering techniques are effective in crime association/

prediction, and social network analysis can facilitate crime association/pattern

visualization.

• Crime Comparison

• Crime Clock

• Outbreaks Detection

• Web Crawling – multi threaded crawlers, preferential crawlers,focused crawlers

• Document Classification – stop words removal,lemmatization/stemming, tf-idf, syntactic and semanticarrangement of words, support vector machines, word net/ ontology,different error costs, sampling techniques, cross validation

• Entity Extraction – sentence splitting, tokenizing, POS tagging,supervised/ semi- supervised/ unsupervised entity extraction

• Duplicate Detection – near duplicate detection, finger prints(shingles, simhash), hamming distance

Preliminaries

The Proposed System

This paper presents a web based crime analysis system.

Sri Lankan English newspapers (Daily Mirror, The Island,

and Ceylon Today) are used as the source for details of

crime incidents.

Newspaper articles are crawled using a focused crawler

and then they are classified.

The Proposed System

Required entities are extracted from classified crime

articles and duplicate detection is performed.

By using these preprocessed data, crime analysis

operations are performed and results are displayed

using a web based GUI.

Unlike most systems, this system is open to anyone who

is interested in crime analysis.

The Proposed System (cont.)

Web CrawlingCrime Analysis and Prediction

Document Classification Entity Extraction Duplicate Detection

Our Solution

• Crawler – crawler4j, Jsoup, cookie handler

• Document Classifier – Weka, LibSVM, SMOTE, Different

Error Costs

• Entity Extractor – GATE, ANNIE, Stanford NLP, Google

Map API

• Duplicate Detector – 64 bit simhash calculator,

murmur hash calculator, hamming distances

• Web Interface – HighCharts, Java scripts, AJAX

Implementation Details

Results

Crime Hot Spot Analysis

Crime Comparison

Crime Pattern Visualization

Conclusion

• The proposed system performs crime analysis

operations such as hotspot detection, crime comparison

and crime pattern visualization.

• Graphical user interface of the system uses graphs and

diagrams to display the results which make crime

analysis a very simple task.

• Therefore law enforcement officers and other

interested users will be able to use this system

effectively and efficiently for crime analysis processes.

• Also this is a publicly accessible system, so that anyone

who is interested in this area will be able to use this

system freely.

Future Work

• Crime prediction is expected to be implemented in future

to enhance the functionality of the system.

• Comprehensiveness of the news article collection can be

further improved by extending the news article crawler to

crawl more news websites.

• Linguistic knowledge (WordNet, Ontology, etc.) can be

incorporated with the document classifier module in

order to improve the accuracy of the classification

process.

• Entity extraction module can be improved by

incorporating more rules which will improve accuracy

and comprehensiveness of the entity extraction process.

Questions?

Thank You!