The Bases of Association Rules of High Confidence

TL;DR

This paper introduces a distributed algorithm for computing high-confidence association rules in binary tables, based on the D-basis method, enabling attribute ranking and tested on transaction and medical datasets.

Contribution

It presents a novel distributed approach for deriving high-confidence association rules from large datasets, extending the D-basis algorithm for practical applications.

Findings

Effective rule extraction demonstrated on real datasets

Algorithm enables attribute ranking based on relevance

Preliminary tests show promising results

Abstract

We develop a new approach for distributed computing of the association rules of high confidence in a binary table. It is derived from the D-basis algorithm in K. Adaricheva and J.B. Nation (TCS 2017), which is performed on multiple sub-tables of a table given by removing several rows at a time. The set of rules is then aggregated using the same approach as the D-basis is retrieved from a larger set of implications. This allows to obtain a basis of association rules of high confidence, which can be used for ranking all attributes of the table with respect to a given fixed attribute using the relevance parameter introduced in K. Adaricheva et al. (Proceedings of ICFCA-2015). This paper focuses on the technical implementation of the new algorithm. Some testing results are performed on transaction data and medical data.