Intelligent Customer Behaviour Analysis in the Norwegian Market

Abstract

Market basket analysis identifies item patterns in data, commonly used in retail to understand customer shopping habits and inform business decisions. Challenges arise with large, high-dimensional datasets. We propose a framework for market basket analysis using dimension reduction and clustering on data from a major Norwegian grocery retailer. This reduces complexity, allowing us to visualize and group data using clustering. The aim is to group similar transactions for association rule mining on a smaller subset. Our research goal is to develop a mobile application for customer grouping and pattern analysis. We apply K-means for grouping and Apriori for rule mining. We evaluate multiple dimension reduction techniques and cluster validation methods. This proved challenging due to dataset intricacies. Results favour t-SNE for dimension reduction, as it effectively separates transactions. Apriori yields many trivial rules, but 'Vegetables/potatoes' emerges as significant. A business case is needed for actionable rules. A better product hierarchy for detailed cluster analysis is also beneficial. Future work should explore improved dimension reduction and clustering assessment methods. The full code can be downloaded from: https://github.com/YousIA/ConsumerAnalytics.