Efficient Association Rule Mining Algorithm, A Model-based Transformation

Mining frequent patterns in transaction databases, time series databases, and many other kinds of databases has been studied popularly in data mining research. Most of the previous studies can be induced to two kinds: One kind adopt an Apriori-like candidate set generation-and-test approach. However, candidate set generation is still costly, especially when there exist long patterns. Another kind adopt an pattern fragment growth method like FP-growth to avoid the costly generation of a large number of candidate sets. FP-growth method is faster than the Apriori algorithm, but still suffer from the cost of frequent pattern generation through analysis on the conditional pattern bases.In this paper,a frequent pattern mining method FP-reduce algorithm based on frequent pattern tree(FP-tree) structure is proposed. It adopt the FP-tree structure to store all the FP information. FP-reduce algorithm can reduce the length of frequent pattern step by step using the left save method i. e. for each item of each itemset, left save method tries to copy the subset of the itemset except the item. Then the subset is added into the FP-tree in order to save the information of the original itemset. Finally we can directly get all the frequent patterns from the FP-tree.Our performance study shows that the FP-reduce algorithm is efficient for mining frequent patterns, and is better than FP-growth in linearity especially when the data set is extended to some extent.