Disk Resident Taxonomy Mining for Large Temporal Datasets

Abstract

Mining patterns under constraints in large data is a significant task to advantage from the multiple uses of the patterns embedded in these data sets. It is obviously a difficult task because of the exponential growth of the search space. Extracting the patterns under various kinds of constraints in such type of data is a challenging research. First, a memory-based, efficient pattern-growth algorithm, Forest Mine, is proposed for mining frequent patterns for the data sets and then consolidating global frequent patterns. For dense data sets, Forest-mine is integrated with FP-Tree dynamically by detecting the swapping condition and constructing FP-trees for efficient mining. Such efforts ensure that forest mine is scalable in both large and medium sized databases and in both sparse and dense data sets.