DATA ANALYSIS TECHNIQUES FOR FRAUD DETECTION

Context : MCA students have two hours a week dedicated for seminar. There are 30 minute talks and 4 minute talks. Following is a summary of a 4 minute talk
Date: 12 - June 2011
Speaker: Abhishek Ramachandran

See Fraud is basically to deprive someone of something, of value for their own benefit. Even fraud is a million dollar business and is increasing every year. A global survey conducted in 2007 stated that almost 50% companies all over world has been victim of fraud. So even though the technologies are increasing every year in the same way the criminals are finding new way to commit fraud. In addition to that business re engineering and downsizing may help in reducing fraud to some extent but new Information Systems may present new opportunities to commit fraud. There has been traditional ways of data analysis to prevent fraud but its very complex & time consuming and one should have knowledge of financial , business practices & law.

Fraud often consists if same instances or incidents like Palshikar said fraud may be of same content and appearance but usually are not identical. The first industries to use the Data analysis techniques are the telecom , insurances and the banking sector. Even retail industries suffer from fraud at POS. So retail industries have started using CCTV cameras in shops. Basically DA techniques are divided into 2 classes i.e Statistical and AI Techniques(artificial intelligence).

Statistical comprise of knowledge Discovery in database and Statistics. Whereas in AI techniques comprise of Data mining and Machine Learning. Statistical Techniques we use are Probability Distribution, averages, quartiles& performance metrics to detect fraud. Even computing the user profiles too.

And in the AI techniques there is Data mining and Machine Learning. Here data mining is used to classify, cluster and segment data to find associated rules and patterns in the data. Machine Learning techniques to automatically identify characteristics of fraud. In the earlier days, data analysis techniques were oriented towards extracting quantitative and statistical data characteristics. They used various data interpretations to get better insights of process behind the data. So to go beyond, the data analysis system has to be equipped with substantial amount if background knowledge and to perform tasks based on that knowledge. So I would like to conclude by saying that reducing fraud is a big ask but not a big task .