Knowledge Detection with Data Mining to Fight Insurance Fraud
The amount of data organizations store is rapidly increasing and we often face the challenge to attain useful information from data that exceeds in volume and complexity. Data mining enables detection of knowledge in data, which is hidden to the human eye. And it is the use of such knowledge in business that sets companies apart from their competition nowadays.
Fraud occurs at different stages. The first fraud opportunity could manifest during a pre-insurance stage when the application for insurance is issued and false data are deliberately used with the intent to defraud the insurance company or at best get a better price. Furthermore, post-insurance stage frauds are also common (see Whitepaper: Crash for Cash Scams). For a competent analysis of potential fraudulent activities we have to analyse all the data we have about the accident, claimant, driver and potential passengers, other people involved in the accident, witnesses, injuries, medical treatments, potential mechanical costs, damage, vehicles involved, etc.
Understanding the problem and the data set is crucial, as well as the stage of data preparation, if we want to effectively apply different data mining methods, followed by a coherent interpretation of results. According to Sithic and Balasubramanian (2013) the following steps are fundamental when applying data mining techniques for insurance fraud detection:
1. Data cleaning: Remove noise and inconsistence in data.
2. Data integration: Combine various data sources.
3. Data selection: When searching through the database we must focus on data related to the selected subject.
4. Data Transformation: The data must be transformed into the form appropriate for mining.
5. Data Mining: Data models extraction must rely on advanced technology and know-how.
6. Pattern evaluation: Our existing knowledge and expert external knowledge should help us evaluate the obtained models.
7. Knowledge presentation: The obtained data must be transformed into useful information. Visual presentation might help.
Technology might assist us during the extraction of new knowledge and rules. Nowadays, insurers are using data mining tools and techniques to examine high volume data set and to find out patterns that can help them to separate anomalies from genuine claims. Start your own fraud detection data mining project and contact us.