Efficient Incremental Pattern Mining from Semi-structured Dataset

摘要

Tree-structured frequent pattern mining is an important issue in semi-structured data mining. In this paper, we study the problem of incremental pattern mining from labeled ordered trees by using the knowledge discovered in the previous mining operations, and present an efficient algorithm IncFreqt for discovering frequent substructures from a large collection of semi-structured dataset. Because the inserted position of the increased data tree is not limited to the root of the original data tree, this paper introduces a new expansion algorithm called Bridging expansion, which is used to find the patterns that can not be discovered by normal rightmost expansion in the case that the new data tree is not inserted to be a subtree of the root of the original one. The experimental results show a great improvement in the efficiency of our algorithm compared with that of non-incremental pattern mining algorithm.