Bayardo described in [4] he performed a version of Max miners found frequent itemsets maximum length. To our knowledge, there is no other available algorithms with I Max itemsets. Many algorithms exist for MFI and mining. Agarwal et al. [1] points out that DepthProject run more than an order of magnitude faster than Max miners. Burdick et al. [6] showed that the MAFIA faster than DepthProject by a factor of 3-5. Grahne and Zhu [10] shows FPMAX performance gain compared to the MAFIA and GenMax. To compare performance, we modify the algorithm a little efficiency FPMAX MAFIA and I frequent itemset maximum length. Why we have chosen these two algorithms? In fact, DepthProject, MAFIA and GenMax share a lot in common: search for items mesh subset (or lexicographic tree) in a way the first worm, apply the lookahead pruning and dynamic rearrangement to reduce the search space, use a compression technique to quickly support counts. So, we picked out the MAFIA as the representatives of the three algorithms. FPMAX is essentially different from the above three and similar to our algorithm, because it not only uses the same tree FP-structure but also expand the FP-growth algorithm. So we also have chosen FPMAX as a competitor. Figure 13 illustrates the results of comparing the three algorithms for mushrooms, chess, Connect4 and Pumsb *, respectively. The left column shows the running time of the algorithm. The x axis is specified the minimum user support, expressed as a percentage, while the y axis shows the running time in seconds. The middle column comparing the number of itemsets candidate handled by FPMAX_LO and LFIMiner. The x axis is the minimal support, and the y axis displays the number. The right column compares the total of all the button FP-tree created by FPMAX_LO and LFIMiner. The x axis is the minimal support, and the y axis displays the number.
đang được dịch, vui lòng đợi..
![](//viimg.ilovetranslation.com/pic/loading_3.gif?v=b9814dd30c1d7c59_8619)