ICIAC 2012 Stream 2: classes and workshops

Abstracts and slides

CLASS 2C: Getting inside the criminals OODA loop - Timely intelligence analysis (G)

JonMichael Teare, Project Manager, CrimeDeter

Slides: yet to be supplied by presenter

Fast analysis of multiple, data heavy sources can break the confidence of the criminal. Criminal decision making like any other decision occurs in a recurring cycle of observe-orient-decide-act. This session will show, real world examples of how fast paced delivery of analysis to law enforcement can help disrupt criminal activity and improve law enforcement planning. The case studies are both pre and post arrest and will cover:

  • Identification of potential handlers helping drive level 1 crime (metal theft SMDs)
  • Correlation of criminal movements with reported crimes utilising various data sources including cell tower, WiFi APs and GPS history data
  • Interview and case preparation
CLASS 2D: Detecting fraudulent activity - an intelligence-led approach (I)

Martin Gill, Director of Intelligence Services, Keoghs LLP

Slides: speaker was ill so did not present

Insurance fraud costs the insurance industry more than £2bn each year - and impacts all policyholders via increased premiums. In this session we consider the nature and scale of the insurance fraud issue and how intelligence enables us to identify, validate, investigate and resolve insurance fraud matters. We will also look at the use of data mining, statistical analysis and predictive analytics in attempting to deny fraudsters access to insurance and prevent fraudulent claims before they happen.

Keoghs were one of the first law firms to develop an insurance fraud capability, and have worked with many of the UK's leading insurers on their counter-fraud responses for more than a decade. The firm have pioneered the use of intelligence as an integral part of a robust counter-fraud strategy and today our 28-strong intelligence team is the largest specialist unit of its type in the insurance industry. By working in partnership with our clients, the police, Insurance Fraud Bureau and other regulatory bodies we delivered £100m in fraud savings in our last financial year.

Insurance fraud costs the insurance industry more than £2bn each year - and impacts all policyholders via increased premiums. In this session we consider the nature and scale of the insurance fraud issue and how intelligence enables us to identify, validate, investigate and resolve insurance fraud matters. We will also look at the use of data mining, statistical analysis and predictive analytics in attempting to deny fraudsters access to insurance and prevent fraudulent claims before they happen. Keoghs were one of the first law firms to develop an insurance fraud capability, and have worked with many of the UK's leading insurers on their counter-fraud responses for more than a decade. The firm have pioneered the use of intelligence as an integral part of a robust counter-fraud strategy and today our 40-strong intelligence team is the largest specialist unit of its type in the insurance industry. By working in partnership with our clients, the police, Insurance Fraud Bureau and other regulatory bodies we delivered £100m in fraud savings in our last financial year.

CLASS 2E: Advanced hotspot analysis - spatial significance mapping using nearest neighbour analysis and the Gi* statistic (A)

Spencer Chainey, Department of Security and Crime Science, University College London

Slides: Chainey

In this class we explore the use of the Gi* statistic as a practical method for incorporating spatial statistical significance testing into hotspot analysis. The workshop begins by briefly reviewing the utility of common hotspot mapping techniques: point  mapping, spatial ellipses, choropleth mapping, and kernel density estimation (KDE). Research shows that KDE is the best of these common techniques, however it is not without its flaws.

We introduce the concept of applying significance testing to spatial data by illustrating the use of the nearest neighbour index for determining if hotspots exist in our data.  We then   explain ithe principles behind the Gi* statistic and its use as a technique for spatial significance mapping. This includes a step by step guide on how to use the Gi* statistic and advice for determining appropriate parameter settings. We show evidence from research that the Gi* statistic goes beyond KDE because it can identify (in map form) those areas where the clustering of crime points is significant. That is, it can determine areas that can be statistically defined as hot from those that are not, plus represent this thematically by the level of statistical significance i.e. 90%, 95%, 99% or 99% confidence levels.  We illustrate the use of the Gi* statistic by using the free Excel-based software, Rook's Case (with outputs being imported into any GIS software), and the functionality in standard ArcGIS. 

Page last modified on 30 jan 13 17:04