Interventions to Detect Insurance Fraud - Literature review Example

Interventions to Detect Insurance Fraud Introduction The issue of fraud and buildup is of major concern to insurance firms. Fraud comprises of illegitimate claims while a buildup is the exaggeration of loss amounts. All these comprise of fraudulent acts that an individual may engage in to defraud an insurance firm. In the recent past, the insurance industry is concerned with the detection and prevention of such fraudulent behaviour. For example, the rising number of fraudulent automobile claims has sparked this major interest in companies (Tennyson and Pau, 2002). An insurance contract usually involves an agreement between the insured and insurer for the insuring firm to cover a specific risk (Tennyson and Pau, 2002). However, the behaviour of the insured, who is to receive compensation in the event of the risk, is not always honest (Tennyson and Pau, 2002). This gap is what many insured individuals use to defraud insurance companies especially automobile insurance companies (Dionne and Laberge-Nadeau, 1999). Most automobile insurance companies are concerned about this vice and are continually using auditing patterns to control the situation (Tennyson and Pau, 2002). Detection of Fraud The detection and prevention of fraud is not an easy task. Many companies spend considerable amounts of time and money to detect and prevent fraud (Schiller, 2002). Some companies, however, use auditing patterns to control fraud (Schiller, 2002). This approach is also known as costly state verification and assumes that the insurer can get information about a claim without incurring an auditing cost (Schiller, 2002). The method implicitly assumes that the audit process, which uses auditing technology, always discerns if a claim is honest or fraudulent (Bolton & Hand, 2002). The main discussion usually concerns the design of the contract which often minimizes the insurer’s cost including the cost of the audit and total claim payment (Schiller, 2002). Models using this framework normally use the amount of the claim to make the decision of applying the monitoring techniques. Other works suggest the comparison of deterministic audits against random audits (Artis et al 2002). However, these auditing techniques that are applied in the detection of fraud by these organisations may not always be perfect and the claims may be misclassified at times (Artis et al 2002). This possibility has made many companies consider more accurate methods of fraud detection and subsequently prevention such as the use of data mining and data matching techniques and fraud detection software. Data Matching Techniques Data matching entails the combining and comparison of data from different sources. Individuals using data matching techniques in insurance firms usually find definitive data sources from the insured customers and use a series of proprietary matching algorithms and methods to detect any cases of fraud from the insured (Olson and Delen, 2008). Data matching techniques covers two related and distinct functions of computerised databases (Olson and Delen, 2008). One of the functions is the comparison of a person’s details with a number of databases. The other is the act of performing a side-by-side comparison of the databases to detect anomalies, trends and potential duplicates (Olson and Delen, 2008). Insurance companies use data matching to detect fraud in three major ways; data matching agents, data sharing and data bureau (Ebbers, 2013). In data sharing, the users exchange the information they possess directly with other users for the purpose of matching (Ebbers, 2013). For instance, an insurance firm can match the information they have from a claimer in the event of an accident with the information that the police have to verify if the individual actually reported the accident. Data matching agents entails the restricted exchange of data between parties while data bureau comprises of the centralization of computer records providing a shared database to users where information can be matched (Ebbers, 2013). Data matching techniques are important methods of detecting fraud that are being used by insurance firms to detect and prevent frauds. In the comparison and matching of data with other firms, insurance companies are taking advantage of a number of tools and technologies (Ebbers, 2013). For instance, insurance companies use neural networks, fuzzy logic, phonetic matching, and intelligent systems to match, compare, and process information (Ebbers, 2013). Data Mining Techniques Data mining techniques are an important tool for insurance companies in the detection of fraud (Perner, 2004). The advancement of computer software technology and gaining popularity of social platforms has provided insurance a plethora of digital inputs that aid in the detection of fraud using analytics (Perner, 2004). Manual handling of fraud has been costly for many insurance companies as these methods lack efficiency (Perner, 2004). With the application of data mining techniques and tools, an insurance firm the number of frauds that go undetected has been reduced significantly (Perner, 2004). The insurance company can use a set of three innovative fraud detection methods. These methods include Social Network Analysis (SNA), predictive Analytics for big data and social customer relationship management (CRM) (Verma and Ramakrishani, 2014). In a SNA, a team of analytics receives data from various sources. The team uses the information and analyzes the risk and likelihood of fraud. Technologies such as sentiment analysis, text mining and content categorization of social network data are integrated into the fraud identification and prevention model (Verma and Ramakrishani, 2014). An alert is generated depending on the score of a particular network. The investigators then leverage this information and engage in deeper research on the claim (Verma and Ramakrishani, 2014). The issues are identified and added to the business case system if proven to be fraudulent forming a section of the hybrid framework. By use of predictive analytics of big data, investigators in insurance firms can detect scam and fraud in the claims of insured individuals (Verma and Ramakrishani, 2014). This is usually achieved by the analysis of unstructured data in reported claims using text mining techniques to identify indicators of fraud (Verma and Ramakrishani, 2014). The system works by identifying clues that may remain hidden in the long reports written by the claim adjusters in an insurance firm (Verma and Ramakrishani, 2014). This is possible as the computing systems are based on business rules and can spot evidence and clues of possible fraud. Psychologists propose that individuals who commit fraud usually alter their story/account over time and these discrepancies are what the fraud detection systems spot. Social customer relationship management is a process that insurance firms employ to detect and further prevent fraud (Verma and Ramakrishani, 2014). Insurance companies that link their CRM to social media can use social CRM. This process uses a listening tool which usually extracts data from the social chatter and social media platforms (Verma and Ramakrishani, 2014). The information stored in the CRM as well as the reference data are fed into a claim management system, which analyzes the data based on the firm’s business rules and then sends a response (Verma and Ramakrishani, 2014). The investigators then confirm whether the information from the claim management system is fraudulent or not as the output from the system acts as an indicator and not a final decision. Diagrammatic representation and further illustration of the methods are provided in the appendix section of this paper. Summary The 21st century business scape calls for the automation of all business processes (Anderson et al 2010). By the use of modern techniques of data mining, data matching and software in the detection and prevention of frauds, insurance firms can reduce the amounts of loss they incur in the payment of claims that are fraudulent. Investigators in these insurance firms can deal with the exploitation weaknesses from individuals. The modernized technique not only aid in the detection and prevention of frauds in insurance firms, but also reduces the compensation time of legitimate claims in the insurance firms. References Anderson, J. M., Heaton, P., & Carroll, S. J. (2010). The U.S. experience with no-fault automobile insurance: a retrospective. Santa Monica, CA, RAND. Artis, M., Ayuso, M. & Guillen, M. (2002). Detection of Automobile Insurance Fraud with Discrete Choice Models and Misclassified Claims. The Journal of Risk and Insurance, Vol. 69, No. 3, pp. 325-340. Retrieved from: http://onlinelibrary.wiley.com/doi/10.1111/1539-6975.00022/abstract?deniedAccessCustomisedMessage=&userIsAuthenticated=false Bolton, R.J. & Hand, D.J. (2002). Statistical Fraud Detection: A Review. Statistical Science, Vol. 17, No. 3, 235–255. Retrieved from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.92.9299&rep=rep1&type=pdf Dionne, G., & Laberge-Nadeau, C. (1999). Automobile insurance: road safety, new drivers, risks, insurance fraud and regulation. Boston [u.a.], Kluwer Acad. Publ. Ebbers, M. (2013). Real-time fraud detection analytics on IBM System Z. [Poughkeepsie, N.Y.], IBM Corp., International Technical Support Organization. http://proquest.safaribooksonline.com/?fpi=0738437638. Olson, D. L., & Delen, D. (2008). Advanced data mining techniques. Berlin, Springer. Perner, P. (2004). Advances in data mining applications in Insurance ; 4th Insurance Conference on Data Mining, ICDM 2004, Leipzig, Germany, July 4-7, 2004 : revised selected papers. Berlin, Springer. Schiller, J. (2002). The Impact of Insurance Fraud Detection Systems. University of Hamburg. Retrieved from: http://onlinelibrary.wiley.com/doi/10.1111/j.1539-6975.2006.00182.x/abstract;jsessionid=1BBC9B03B86CCFE2706870AE1CCF41B7.f02t03?deniedAccessCustomisedMessage=&userIsAuthenticated=false Tennyson, S. & Pau, S. (2002). Claims Auditing in Automobile Insurance: Fraud Detection and Deterrence Objectives. The Journal of Risk and Insurance, Vol. 69, No. 3, pp. 289-308. Retrieved from: http://www.jstor.org/stable/1558679 Verma, R. & Ramakrishani, S. M. (2014). Using Analytics for Insurance Fraud Detection: 3 Innovative Methods and a 10-Step Approach to Kick Start your Initiative. http://www.infosys.com/FINsights/Documents/pdf/issue10/insurance-fraud-detection.pdf Appendix Figure 1: Social Network Analysis (SNA) Source: http://www.infosys.com/FINsights/Documents/pdf/issue10/insurance-fraud-detection.pdf Figure 2: Processes in Big Data Analysis Source: http://www.infosys.com/FINsights/Documents/pdf/issue10/insurance-fraud-detection.pdf Figure 3: Steps in Fraud Detection using Big Data Analysis Source: http://www.infosys.com/FINsights/Documents/pdf/issue10/insurance-fraud-detection.pdf Figure 4: Insurance Fraud Detection Using Social CRM Source: http://www.infosys.com/FINsights/Documents/pdf/issue10/insurance-fraud-detection.pdf Read More