| Summary: | 碩士 === 中華大學 === 資訊管理學系 === 93 === The data mining technology has already been widely applied various kinds of fields. However, with the application of the database and influence of the global informationization, the amount of data increasing explosively, the traditional data mining technology was unable to support so heavy data amount gradually, and it can not satisfied the user’s request for efficiency. Therefore, this thesis proposes that a parallel distributed mining algorithm based on FP-Tree structure, utilize the characteristic that FP-Tree structure can be disassembled, the loading balanced by exchange the tree structure’s information to each other in the parallel operation. In addition, a simple and trusty calculates formula in load of degree, carry out the load evaluation to the item that the processors responsible for. With the low exchange times and the mechanisms of holding some item, to decreasing of the data transmission quantity between the processors and the time that processors dealing the data of the transmission processes. To compare with the Parallel frequent pattern tree algorithm (PFP-Tree), with the PC Cluster of 16 processors, our algorithm could be provided the 18% rate of improving efficiency. Therefore, the proposed algorithm could utilize the resources indeed and accelerate the mining speed effectively.
|
|---|
