The Bank On The Net Data Stream Based On Sliding Window Frequent Pattern Study

At present, E-bank, as the virtual bank counter on the Internet, is taking the place of the traditional bank service in counter. E-bank provide customers with personalized service, not only the traditional service but the developing service through information thechnology. E-bank helps the commercial bank cutting costs and enhancing profitability. E-bank provide customers with numerous and personalize service and break through time and space with information thechnology.It is a key for E-bank to obtain core competitiveness from vast explosive growthed and dynamic changing data.It is on the priority to consider how to use the finite system resource to obtain the useful information from fast analyzing and managing vast data. Data is changing, frequent, unpredictable and reaching with datastream. This pattern of data is not a lasting relation table modeling but a instantaneous data stream modeling. Researching the data steam frequent pattern in e-bank can help e-bank make better activity decision.This paper firstly study the data stream mining modeling and relevance theory, concluding the present results in this field and the concept of e-bank data stream. Then the research analyze the data stream mining at present to put forward improved mining maximal frequent patterns over data streams algorithm via sliding window, BFPM-Stream. It used transaction and time sensitive sliding window to resolve the unfavorable effects caused by the unsteady speeds of data stream. According to the e-bank maintainers’needs that it is impossible to give an appropriate value for the minimal support threshold that is used to mine a data stream. This paper put forward a method for mining the Top-K frequent patterns in a sliding window. Finally, the study make an experiment on the two algorithms with simulated data. The experimental results conclude that the algorithms have well-performance, stability and practical significance in E-bank.