This study aims to examine brand-level co-occurrence structures in retail transaction data using the FP-Growth algorithm. It explores how validated association-rule outputs may support enterprise decision-making by serving as interpretable decision-support artefacts within enterprise information management environments.
The FP-Growth algorithm was applied to approximately 28,000 transactions collected from 2 supermarket branches in distinct retail contexts (student-area and city-centre) in Denizli, Turkiye. Association rules were evaluated using support, confidence and lift metrics and further validated through threshold-sensitivity analysis and cross-store stability testing to assess the consistency of the extracted association rules across parameter settings and retail contexts.
The findings reveal context-specific brand co-occurrence patterns, with national-brand clustering dominating in the student-area store and hybrid national–private label configurations emerging in the city-centre store. Several association rules remained stable across parameter settings and retail contexts, indicating their potential use as interpretable decision-support inputs for retail analysis.
Rather than proposing a new mining algorithm, this study contributes by demonstrating how validated association-rule outputs can be interpreted as decision-support artefacts within enterprise information management. It introduces context-sensitive rule stability as a validation approach and presents a conceptual framework illustrating how FP-Growth outputs may support DSS and business intelligence environments.
