Single-dimensional Association Rule Mining Algorithm

Association rules mining, as the most important subject in data mining, reveals the corelations between itemsets and therefore can be widely applied to many fields such as market basket analysis, corelation analysis, classification, web-customised service, etc. Since 1993 when R. Agrawal, R. Srikant firstly proposed the concept of association rules, a lot of algorithms have come up in mining of association rules. While most of these algorithms are based on Apriori line, will generate a huge number of candidate itemsets, need multiple scans over database, and maintain a big hash tree, so the time and space complexity is too high.This paper proposed the constrained-tree based algorithm Fpmine for fast mining of frequent patterns, and the association rules mining algorithm SPF based on threaded frequent pattern tree. The two algorithms dwell on the single-dimension association rules mining in database. Experiments show that Fpmine runs two times as fast as the most recently proposed algorithm Fp-growth and saves half the memory; moreover, our algorithm has a quite good time and space scalability with the number of transactions, and has an excellent performance in dense database mining as well; SPF has excellent time and space efficiency; Fpmine-SPF algorithm has a far taster speed in association rules mining than the widely used Apriori algorithm and has wonderful scalability.Association rules mining always generates too many rules, which makes ii difficult to pick the valuable rules where from. In order to tackle this problem, recent studies raised an effective alternative, mining frequent closed itemsets rules. Frequent closed itemsets rules imply all the association rules but greatly reduce the rules number. Mining of frequent closed itemsets greatly improve the rules mining effiency and effectivity of the resulting rules, hence release the users of the burden.We implement the frequent closeu itemsets mining algorithm FCIS and the frequent closed itemsets rules mining algorithm CI_RULES. Experiments diplay that compared with the latest published algorithm-the CLOSET algorithm proposed by Jiawei Han, our algorithm is more than two malgtitudes faster and has fabulous scalability; CI_RULES algorithm overtakes Fpmine-SPF algorithm by more than two malgtitudes and has extraordinary scalablity.