Performance analysis on association rule in data mining

One of the most important problems in data mining is to find association rules. The association rule mining can be classified into two main categories: the level-wise algorithms and the tree based algorithms. The level-wise algorithm like Apriori, scan the entire database multiple times and also generate a huge number of candidate sets. It also needs to repeatedly scan the database and check a large set of candidates by pattern matching. Tree based algorithms, like FP-tree, scan the database only twice. One scan may be needed for FP-tree construction and another scan for adding new items into the tree. But it takes more time when new data are added to an existing database. Another tree based algorithm, P-Tree is constructed by a single scan of a database and it updates the P-tree by one scan of new data. The above said three algorithms are implemented by using C++/java/.net and their performances are evaluated by using synthetic dataset with respect to number of scanning of dataset. The performance study shows that in majority of cases, Pattern Tree achieves better performance and efficiency than Apriori and FP algorithms.