Unsupervised profiling methods for fraud detection

DATA MINING METHODS IN FRAUD DETECTION

DATA MINING METHODS IN FRAUD DETECTION 2

Data Mining Methods in Fraud Detection

Name;

Mohamed Farah Shidane

Student Number;

20161105024

Title

Application of data mining methods in fraud detection

Objectives

Nowadays there has been a significant increase in the number of fraudulent cases. This has led to an increase in loss of billions of dollars worldwide annually. The biggest setback as a result of fraud includes transactions such as the purchases on the internet, transactions involving mail as well as telephonic transactions. These involving credit and debit cards as well as other modes of commerce. In many instances, the people involved in carrying out these kinds of fraud have significant knowledge on how to go around the system (Kirkos et al., 2013). Therefore in most cases, the fraud detection is done after the fraud has been committed. Institutions are therefore required to have various fraud detection models put in place to deal with such incidences now and in the future. Data mining, therefore, can just be said to be a tool to discover a given pattern, an outlier or even any anomalies.

The scope of the Research

This research basically shows the application of data mining techniques that are quite instrumental in detecting cases of fraud through the concept of data mining. It may also focus on some of the recent methods that are being employed in credit card, in telecommunication as well as in computer intrusion and fraud detection.

Literature Review

The most common kinds of fraud classification include financial fraud that may generally be described as a kind of theft through either an individual or an institution. This may involvetaking money illegally and without the consent of the owner. The complex nature of the current economy posits that different kinds of fraud are likely to occur. They include; bank fraud, securities fraud, occupational fraud, and taxpayer fraud. Money laundering can also be termed as a type of fraud since it seeks to gain through dubious channels.Computer intrusion may be termed as fraud involving the compromise of a computer system by bypassing its passwords and making it enter into an insecure state (Chan et al., 1999).

This kind of intrusion may be detected through misuse detection systems. The systems have the ability to match the current activities with attacks that may have happened previously on their database. An anomaly detection system also has the ability to comprehend and understand the normal activity of a computer and single out any activity that deviates from the computer’s normal patterns (Phua et al., 2010).

There is also insurance fraud that normally happens when the insurance company deceives people all with the aim of benefiting from their money. This may be done through inflated premiums of failing to pay valid claims.(Kirkos et al., 2013). Insurance fraud normally happens when a client puts wrong information willingly on the insurance form with the aim of deceiving the company or even an insurance company may omit crucial information in an insurance claim.

Data Mining Techniques in Fraud Detection

All kinds of data mining techniques that may be used can be categorized under supervised and unsupervised learning. Through the use of supervised learning, data collected may be either fraudulent or not whereas unsupervised learning normally fails to make use of labeled records. As discussed above credit card fraud detection systems are usually used but in most cases they are confidential and the information is not given to the general public. By using the outlier detection technique, an outlier which is generally an observation that deviates from the expected observation. This will occur in a manner that will raise suspicion that it was generated differently and not as it was required.

A neural network is another data mining technique that is basically a set of interconnected nodes that have been designed so as to copy how the human brain functions. These neural networks can be formulated for use in supervised or unsupervised learning. Expert system is a technique that may be defined as a computing system that has the innate ability to reason about fraudulent practices and give best ideas of how to solve the problem. These systems internalize the knowledge about attacks that are likely to occur. They then perform real-time monitoring of all activities carried out by users (Bolton& Hand, 2001).

The model-based reasoning is another technique that is able to detect attacks mainly through visible activities that gather an attack signature. Telecommunication fraud detection techniques generally use the call detail record so as to come up with a complete behavior pattern and their different profiles. This then enables the detection of deviation of any customer that may occur at any given time (Wang, 2010). Visualization method is techniques that generally relies on human pattern recognition so as to detect any kind of anomalies and areprovided with almost close real-time data feeds. The technique combines human detection with machines holding high computational capacity that has the ability to manipulate the number of calls made internationally in a graphical manner. This enables the detection of international calling fraud (Lee et al., 2001).

According to Bolton& Hand (2002), data mining approaches can be used for detecting intrusion. These may enable coming up with new models that have the capacity of detecting new attacks before they can be realized by human experts.

References

Bolton, R. J., & Hand, D. J. (2001). Unsupervised profiling methods for fraud detection. Credit Scoring and Credit Control VII, 235-255.

Bolton, R. J., & Hand, D. J. (2002). Statistical fraud detection: A review. Statistical science, 235- 249.

Chan, P. K., Fan, W., Prodromidis, A. L., & Stolfo, S. J. (1999). Distributed data mining in credit card fraud detection. IEEE Intelligent Systems and Their Applications, 14(6), 67-74.

Kirkos, E., Spathis, C., & Manolopoulos, Y. (2007). Data mining techniques for the detection of fraudulent financial statements. Expert systems with applications, 32(4), 995-1003.

Lee, W., Stolfo, S. J., Chan, P. K., Eskin, E., Fan, W., Miller, M., ... & Zhang, J. (2001). Real time data mining-based intrusion detection. In DARPA Information Survivability Conference & Exposition II, 2001. DISCEX'01. Proceedings (Vol. 1, pp. 89-100). IEEE.

Phua, C., Lee, V., Smith, K., & Gayler, R. (2010). A comprehensive survey of data mining-based fraud detection research. arXiv preprint arXiv:1009.6119.

Wang, S. (2010, May). A comprehensive survey of data mining-based accounting-fraud detection research. In Intelligent Computation Technology and Automation (ICICTA), 2010 International Conference on (Vol. 1, pp. 50-53). IEEE.