Title :
Accelerating Intersection Computation in Frequent Itemset Mining with FPGA
Author :
Shaobo Shi ; Yue Qi ; Qin Wang
Author_Institution :
Sch. of Comput. & Commun. Eng., Univ. of Sci. & Technol. Beijing, Beijing, China
Abstract :
Frequent item set mining is an important researching area in data mining and Eclat is a typical and high performance frequent item set mining algorithm. However, the large numbers of sorted-set intersection computation in the algorithm limit the performance of the algorithm seriously. FPGA is a low-power and high-performance computing platform that has been applied to accelerate parallel data mining successfully. To deal with the problem of the large number intersection computation in Eclat, this paper proposed a FPGA solution to accelerate the intersection computation. And a full comparison matrix structure is provided to perform the parallel intersection computation. And a hardware hash table method is also proposed to eliminate the data dependency in the intersection computation. The experiment results show that our solution can achieve a speedup of 26.7x on intersection computation comparing to the best software implementation existed, and the hardware hash table method can achieve a speedup of 103x. The full comparrison matrix have a better scalability, thus the entire running time of the Eclat algorithm can be decreased extremely.
Keywords :
data mining; field programmable gate arrays; matrix algebra; parallel processing; Eclat algorithm; FPGA; data dependency; full comparison matrix structure; hardware hash table method; high performance frequent item set mining algorithm; high-performance computing platform; intersection computation acceleration; low-power computing platform; parallel data mining; parallel intersection computation; sorted-set intersection computation; Acceleration; Algorithm design and analysis; Data mining; Field programmable gate arrays; Hardware; Indexes; Itemsets; FPGA acceleration; Frequent itemset mining; Full Comparison Matrix; Hardware Hashtable; Parallel Data mining; Sorted-set intersection;
Conference_Titel :
High Performance Computing and Communications & 2013 IEEE International Conference on Embedded and Ubiquitous Computing (HPCC_EUC), 2013 IEEE 10th International Conference on
Conference_Location :
Zhangjiajie
DOI :
10.1109/HPCC.and.EUC.2013.98