Link Analysis: Too Much GUI?
On Using Link Analysis for Fraud Detection: Most link analysis tools offer a sophisticated graphical user interface (GUI), showing connections that otherwise would not be detected. The GUI approach is useful for drill-downs on already known entities or bad actors, but sifting through mountains of GUI graphs can be very time consuming when conducting proactive data mining. It seems most every company offers some kind of link analysis tool with a fancy GUI: InfoGlide, Detica NetReveal, I2 Analyst’s Notebook, to name a few (see full list below). These tools are very powerful and come with a hefty price tag, mostly because of the GUI interface. There is a place for graphical user interfaces, but as a corporation, you must ask yourself: does the need justify the price tag?
I have worked with SAS software for 24 years. During that time, I have developed a series of fuzzy matching protocols, matching algorithms, entity resolution, and link analysis software. Link analysis is definitely available without the fancy bells and whistles. I will also maintain that any system that purports to import your data and link it all together immediately and seamlessly is trying to sell you beach property in Alaska. Every job is customized. The customization may occur through lengthy and expensive configuration fees after purchasing. Did you know, the piece of code that actually performs link analysis can fit on one page? Not all of the fuzzy matching that goes before it, but the actually record linkage mechanism is only one page of code. It links together entity A to entity Z even if they are only related via a common entity “M”, but 20 degrees of separation. Here is a simple example:
Using your own eye, you can tell that the first 3 rows are linked via address even though the addresses differ slightly. Rows D and E are related via address, and rows F and G are related via similar phone number. If family members purchase cell phone service at the same time, they are often given sequential cell phone numbers. This scenario could be an actual case where Dr. Abdel is the founder of “Happy Home Care”, a fictitious home health care clinic. Walker Associates is a legitimate business that is fronting for Happy Home Care, and owns suites A and B in the same building. John Walker owns Walker Associates (as evidenced by the same email), and their daughter is married to Mohammed Abdel, who lives in the basement apartment of 237 S. Ridgeway, which is a single family home.
If you run this scenario via a graphical user interface, you will probably receive a picture that looks like this:
This is great information and shows the linkages. But there could be 1,000,000 legitimate linkages just like this one, showing family relationships, phone numbers and shared bank accounts. The way to distinguish a “good” cluster from a “bad” cluster is to marry it with other meaningful data, such as auto ownership, criminal records, purchase of prepaid phone cards, bankruptcy filings, and more. To show all of this on a screen would be difficult. But flagging this information WITHIN the data itself is easy. You can customize the link analysis within the data to only flag links to someone who had a bankruptcy filing. Or specify that you only want to see clusters where one member of the cluster lives at a certain location. This can all be done with the GUI but you may be still searching for a needle in a haystack if you are sifting through mountains of graphs. By controlling what you see within the data, you can more quickly hone in on aberrancies. You may only be interested in large clusters, so you can easily add a column to calculate the number of members in a cluster/family.
That being said, link analysis with advanced GUIs are not a bad investment. You will still find results. My point is that you don’t need to spend millions of dollars to do it, and you don’t need a sophisticated GUI unless you need to show the linkages to third parties.
If you are interested in learning more about link analysis, you can read up on the following software:
List of Link Analysis Software Tools for Fraud Detection:
- Alaric Systems “Fractals” card fraud detection and prevention systems using proprietary inference techniques based on Bayesian methods.
- Analyst’s Notebook 6, from IBM, conducts sophisticated link analysis, timeline analysis and data visualization for complex investigations.
- Apollo Wipro fraud detection platform, leveraging Big Data and machine learning, with pre-built models for procurement, payments, workforce policy compliance, resource management, and regulatory & contractual violations.
- Aptelisense Compliance Automation Server, advanced real-time fraud prevention and data compliance that requires zero change to applications or systems.
- ArcSight AntiFraud Accelerator Solution.
- Austin Logistics FraudAlert, solutions for collections, marketing, and risk management for consumer credit and Internet transactions.
- Business Data Miners builds highly effective data-driven models and rules to mitigate credit risk and fraud losses; saved its clients over $100 million in the past 2 years.
- Centrifuge, offers analysts and investigators an integrated suite of capabilities that can help them rapidly understand and glean insight from new data sources.
- CPA Detective offers a sophisticated real-time fraud detection solution that evaluates the digital forensics of each visitor and returns the probability of fraud before buying, selling, or fulfilling on a lead or sale.
- CSC Fraud Analytics Suite, combines predictive data modeling technology, identity search technologies, fraud indicator business rules, company claims information and industry data sources to help flag suspicious claims as early as the first notice of loss.
- Dinkla Artificial Intelligence and Fraud Detection pages
- Equifax Fraudscan detects, validates, and verifies potentially fraudulent information automatically and simultaneously at the time of application.
- Equivax Gemini Verify Score Equifax’s credit data with HNC’s fraud-control program to offer an identity authentication score.
- FICO, (formerly Fair, Isaac), offering Falcon and other tools for risk management systems, including credit card fraud detection.
- Fraud Wiki, Fraud Detection and Prevention Wiki.
- FraudBreaker, web based fraud detection software that captures your transaction data and performs real time checks on a wide range of risk factors.
- Friss Fraud Solutions, the leader for fraud and risk detection and settlement in Netherlands; Delivered with best practice fraud indicators en standard interfaces.
- IBM Counter Fraud Management, a holistic, next generation approach to pro-actively preventing, detecting and investigating fraud, financial crimes and improper payments by integrating advanced analytics into a single solution.
- IDES Technologies, a global provider of fraud detection products to solutions providers and transaction processors, mainly for the financial services market.
- InferX, remote data mining solutions for law enforcement, intrusion detection, and related applications.
- Infoglide, developers of patented Similarity Search Engine for finding fraud in the insurance industry.
- IntegraAnalytics from Red Flag Group, a powerful analytics engine that scours multiple systems, ERPs, CRMs, and databases to identify suspicious transactions in real time.
- International Compliance Association, supports and educates compliance professionals in the fight against terrorist financing, corruption, money laundering and financial crime.
- The Lavastorm Analytics Engine (LAE) is designed to augment your current Fraud Management Solution, giving you the ability to keep pace with the rapidly changing fraud environment, while minimizing your costs and delivering rapid ROI.
- LEADMiner, a refinement of Numerically Integrated Profiling System (NIPS), developed for US government fraud detection and trade analysis.
- Magnify PATTERN:Detect™ for uncovering fraud and anomalies, such as fraudulent credit card transactions or network intrusions.
- Nestor, offering risk management products, including credit card fraud detection
- Neural Technologies Decider™, a suite of solutions for the finance industry for advanced modelling and scorecard development for detecting bad debt and application fraud
- NORA™ (Non-Obvious Relationship Awareness™), identifies potentially alarming non-obvious relationships among and between individuals and companies
- Oscar Kilo’s Detect, provides both a rule-based and statistical risk-engine, with applications to Credit Card fraud detection, accounting fraud, and more.
- Plug&Score, scorecard development software that can be used with any phase in the loan cycle from loan origination to fraud detection and prevention
- PredPol, provides targeted, real-time crime prediction designed for and successfully tested by officers in the field.
- RiskShield from INFORM GmbH offers a flexible and highly configurable risk assessment and fraud prevention solutions for payment, card processing, and insurance.
- RootStream Detect, accounting errors and fraud detection software to make internal audits and assessments in accounting databases.
- SAS® Security Intelligence, an enterprise approach to fraud, compliance and security issues
- Searchspace, offers iTM™ – intelligent Transaction Monitor for fraud detection and more.
- Statsoft Solutions for Fraud Detection, focus on medical insurance fraud, Medicare fraud.
- Svivot SN-Sphere™, for creating effective intelligence related to networks (associations of people and organizations working together in a particular context).
- The Modeling Agency, offers PiCard(SM) Intelligent Procurement Card Monitoring System designed to actively detect misuse, target auditing team effort, reduce risk when moving from a purchase order System to credit cards, promote higher purchase card volume sooner, and forecast usage trends.
- StatConsulting, fraud detection based on customer behavior modeling using latest data mining methods together with traditional statistics.
- Xanalys, offering investigative solutions for fraud detection, law enforcement, intelligence, insurance, and more.
- Xtract Fraud Detector, uses adaptive neural nets to analyze customer behavior and detect insurance claims fraud, payment card fraud, and more.
- Wizrule and WizWhy find unexpected rules in data and other applications for fraud detection.
Comments are closed