T. Fawcett and F. Provost. Combining data mining and machine learning for e#ective user profiling. In Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining, pages 8--13, Portland, OR, August 1996. AAAI Press.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Cellular fraud detection is similar to intrusion detection in that unusual behavior is to be distinguished from typical behavior. Historically, manual and ad hoc approaches were used to decide which aspects (features) of customers behavior should be profiled to construct fraud detectors. In [ Fawcett and Provost, 1996 ] an 27 innovative framework utilizing data mining techniques was proposed to automate the fraud detector construction process. First, a data mining program is used to discover the general patterns (rule sets) of fraudulent usage from a large database of cellular calls. Next, these patterns are ....

....of connections in the past 30 seconds, need to be considered as additional features. Traditional feature selection techniques, as discussed in the machine learning literature, cannot be directly applied here since they don t consider (across record boundary) sequential correlation of features. In [ Fawcett and Provost, 1996 ] Fawcett and Provost presented some very interesting ideas on automatic selection of features for a cellular phone fraud detector. An important assumption in that work is that there are some general patterns of fraudulent usage for the entire customer population, and individual 61 customers di#er ....

T. Fawcett and F. Provost. Combining data mining and machine learning for e#ective user profiling. In Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining, pages 8--13, Portland, OR, August 1996. AAAI Press.

....and statistical measures as additional features into the connection records. Traditional feature selection techniques, as discussed in the machine learning literature, cannot be directly applied here since they don t consider (across record boundary) sequential correlation of features. In [FP96] Fawcett and Provost presented some very interesting ideas on automatic selection of features for a cellular phone fraud detector. An important assumption in that work is that there are some general patterns of fraudulent usage for the entire customer population, and individual customers di#er in ....

T. Fawcett and F. Provost. Combining data mining and machine learning for e#ective user profiling. In Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining, pages 8--13, Portland, OR, August 1996. AAAI Press.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC