Dynamic Suspicious Financial Transactions Recognition Based on Multi-dimension Analysis of Data Stream

The limitation in real-time and coverage is the main problem that troubles the suspicious financial transactions recognition in our country at present, and dynamic suspicious financial transactions recognition is an effective way to improve it. To implement dynamic recognition, an algorithm recognizing the suspicious mutation characteristics in financial transaction data stream based on multi-dimension analysis of data stream is proposed in this paper. In the algorithm, according to the features of financial transaction data stream, MCTF (Mutation Comparing Time Frame) is used to help compute the measuring parameters and mutation comparing parameters from different dimensions and concept layers on basis of choosing key attributes of transaction records, constructing structure of data stream cube, and determining general paths. Based on the results computed, the suspicious mutation characteristics in financial transaction data stream can be recognized. By experiments, the algorithm is shown to be able to process financial transaction data stream of huge scale in time in limited storage space, and the processing result can reflect the mutations of frequency, amount, and types of transaction records effectively to help recognize the suspicious financial transactions dynamically.