An Algorithm Of Data Mining Based On Global Frequent Patterns And Implementation Of A Data Mining System

On the background of global integration, business dealings of all trades become unprecedented frequently. Moreover, the growth frequency of information exchange appears fiercely. The character of data source has changed from single environment, small quantity of data and static stored method to distributed environment, flowing data form and dynamic process. How to extract value information from a steady flow of dynamic data in terms of limited software and hardware is one of the hot research topics.Data flow frequent pattern mining turns new features in distributed environment. Single site no longer could offer comfortable condition to data pattern mining which is real-time produced recently. Traditional frequent pattern mining of data flow in distributed environment reserve a huge number of candidates which cause a great deal of memory and computing loss. Further more, high network communication cost inflicts a bad result of low level of resources utility. It is on a high level of pattern mining, basing on the candidates who were produced by improved frequent pattern mining is controlled in certain limits. Mean while, it can be sure that the correction could not be guaranteed unless some suitable condition had been taken for candidates filtering.In distributed environment, the method of data flow global frequent pattern mining is able to distribute mining affairs to local sit which depend on its computing power. And it can reduce the data storage space through modifying the skeleton of FP-tree. Besides that, it could low down the quantity of sending constraint pattern message taking the advantage of data fusion in network communication technique after one time mining. At the same time, it can alleviate the communication pressure of pattern frequent updating relied on dynamic monitoring by information exchange between front and back engine.