A constraint based interactive frequent pattern mining algorithm for large databases.

摘要

Over the past decade, many frequent-pattern mining algorithms have been developed. However, many of them rely on the availability of large memory. Their performance degrades if the available memory is limited because of the overhead and extra I/O costs. Moreover, among the algorithms that mine large databases, many of them do not provide users control over the mining process through the use of constraints. Constraint based mining is very important because it encourages users focus on only those patterns that are interesting to the users. Furthermore, among the algorithms that handle user constraints, many of them do not allow users to interactively change the mining parameters during the mining process. As mining is usually an iterative process, it is important to have an algorithm that supports constraint based mining and allows users to interactively mine large databases.; In this thesis, we design and implement a constraint based interactive mining algorithm, named Inverted Matrix++, that uses a disk based data structure called inverted matrix for mining frequent patterns from large databases and constructs a conditional tree called COFI*-tree for each frequent item from the inverted matrix. Our algorithm facilitates constraint based mining and interactive mining from large databases. Experimental results show the efficiency of our algorithm in constrained interactive mining from large databases.