OInduced: An Efficient Algorithm for Mining Induced Patterns From Rooted Ordered Trees

摘要

Frequent tree patterns have many practical applications in different domains, such as Extensible Markup Language mining, Web usage analysis, etc. In this paper, we present OInduced , which is a novel and efficient algorithm for finding frequent ordered induced tree patterns. OInduced uses a breadth-first candidate generation method and improves it by means of an indexing scheme. We also introduce frequency counting using tree encoding. For this purpose, we present two novel tree encodings, namely, m-coding and cm-coding, and show how they can restrict nodes of input trees and compute frequencies of generated candidates. We perform extensive experiments on both real and synthetic data sets to show the efficiency and scalability of OInduced.