A Parallel Design of Frequent Item-sets Mining on CPU-MIC Heterogeneous Systems's dataset.

With the rapid growth of the text mining technology, knowledge discovery in text is appealing more and more to researchers. Furthermore, association rule mining (ARM), an essential issue in data mining, is widely used in the field of text information. However, due to the exploding amount of information in different fields, the serial ARM algorithm has left much to be desired. In this paper, we propose a new parallel ARM algorithm on corpus, using MPI programming interface. The algorithm mainly makes use of distributed inverted hash index, a communication pattern based on chessboard decomposition and a multi-thread candidate item-set filtering method running on MIC platform to accelerate the generation of frequent item sets.