Market basket analysis: Complementing association rules with minimum spanning trees

Mauricio A. Valle, Gonzalo A. Ruz, Rodrigo Morrás

Research output: Contribution to journalArticlepeer-review

34 Scopus citations

Abstract

This study proposes a methodology for market basket analysis based on minimum spanning trees, which complements the search for significant association rules among the vast set of rules that usually characterize such an analysis. Thanks to the hierarchical tree structure of the subdominant ultrametric distances of the MST, the association network allows us to find strong interdependencies between products in the same category, and to find products that serve as accesses or bridges to a set of other products with a high correlation among themselves. One relevant aspect of this graph-based methodology is the ease with which pairs and groups of products susceptible to carrying out marketing actions can be identified. The application of our methodology to a real transactional database succeeded in: 1. revealing product interdependencies with the greatest strengths, 2. revealing products of high importance with access to another product set, 3. determining high quality association rules, and 4. detect clusters and taxonomic relations among supermarket subcategories. This is highly beneficial for a retail manager or for a retail analyst who must propose different promotion and offer activities in order to maximize the sales volume and increase the effectiveness of promotion campaigns.

Original languageEnglish
Pages (from-to)146-162
Number of pages17
JournalExpert Systems with Applications
Volume97
DOIs
StatePublished - 1 May 2018

Keywords

  • Association rules
  • Market basket analysis
  • Minimum spanning tree
  • Network of products

Fingerprint

Dive into the research topics of 'Market basket analysis: Complementing association rules with minimum spanning trees'. Together they form a unique fingerprint.

Cite this