Discovering Closed and Maximal Embedded Patterns from Large Tree Data

TL;DR

This paper introduces an efficient algorithm for mining closed and maximal embedded tree patterns from large data trees, significantly reducing pattern sets and improving runtime performance.

Contribution

It presents the closedEmbTM-prune algorithm that combines local closedness checking with proactive pruning to enhance embedded pattern mining.

Findings

Generates smaller pattern sets on dense datasets.

Runs significantly faster than previous embedded pattern miners.

Produces complete sets of closed and maximal patterns.

Abstract

We address the problem of summarizing embedded tree patterns extracted from large data trees. We do so by defining and mining closed and maximal embedded unordered tree patterns from a single large data tree. We design an embedded frequent pattern mining algorithm extended with a local closedness checking technique. This algorithm is called {\em closedEmbTM-prune} as it eagerly eliminates non-closed patterns. To mitigate the generation of intermediate patterns, we devise pattern search space pruning rules to proactively detect and prune branches in the pattern search space which do not correspond to closed patterns. The pruning rules are accommodated into the extended embedded pattern miner to produce a new algorithm, called {\em closedEmbTM-prune}, for mining all the closed and maximal embedded frequent patterns from large data trees. Our extensive experiments on synthetic and real large-tree datasets demonstrate that, on dense datasets, {\em closedEmbTM-prune} not only generates a complete closed and maximal pattern set which is substantially smaller than that generated by the embedded pattern miner, but also runs much faster with negligible overhead on pattern pruning.