Efficient adaptive frequent pattern mining techniques for market analysis in sequential and parallel systems

Abstract EN

The classical applications of Association Rule Mining (ARM) are market analysis, network traffic analysis, and web log analysis where strategic decisions are made by analyzing the frequent itemsets from a large pool of data.

Datasets in such domains are constantly updated and as they require an efficient Frequent Pattern Mining (FPM) algorithm which is capable of extracting the required information.

Several incremental algorithms have been proposed to generate frequent patterns, but they are ineffective with very large datasets and do not provide the user interaction to adjust the minimum support value.

This paper first presents an efficient interactive sequential FPM algorithm that uses the knowledge gained in the previous mining steps to incrementally mine the updated database with fewer complexities.

Then to further reduce the time complexity it proposes an efficient interactive and incremental parallel mining algorithm.

It also prepares incremental frequent patterns, without generating local frequent itemsets with less communication and synchronization overheads.