We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Parallel Dynamic Fraud Detection on Market Basket Data.
- Authors
Baoying Wang; Aijuan Dong
- Abstract
Fraud detection is one of the important data mining areas. This paper presents a parallel dynamic fraud detection (PDFD) method on market basket data. PDFD consists of two phases: weighted affinity measure clustering (WC clustering) and fraud detection. First, the data set is analyzed so that items are grouped into clusters in WC clustering phase. Then, each newly arrived transaction is examined against the item clusters retrieved from WC clustering phase. Phase two decides whether the new transaction is a fraud (an outlier). After a period of time, the newly collected transactions are analyzed using WC clustering to produce an updated set of clusters, against which transactions arrived afterwards are examined. The process is carried out dynamically and incrementally. PDFD is implemented using MPI (message passing interface) on parallel machines to achieve high performance. Load balance and the tradeoff between communication time and CPU time are considered. Experiments show that PDFD is faster than the existing methods with a very competitive accuracy. Our future work is to study the optimal number of processors needed for cost effective purpose.
- Subjects
DATA mining; FRAUD; MARKET basket analysis; PARALLEL computers; COMPUTERS
- Publication
International Journal for Computers & Their Applications, 2013, Vol 20, Issue 1, p2
- ISSN
1076-5204
- Publication type
Article