A Shapley-value Index for Market Basket Analysis: Weighting Shapley’s Value

Abstract

ABSTRACT The SIMBA (Shapley value Index for Market Basket Analysis) approach aims to calculate the average revenue generated by an item, including additional revenue from items commonly purchased alongside it. Real-world experimentation in SMCG (Slow Moving Consumer Goods) environments has highlighted the limitations of this method. A novel approach called SIMBAW (Weighted SIMBA) is introduced in this paper, which uses an adapted Shapley value to overcome the aforementioned limitation. The SIMBAW approach weighs items’ Shapley values, and therefore their total attributed values, based on the frequency at which they are the most expensive purchase in their basket and therefore how likely they are to drive their basket in a SMCG scenario. The effectiveness of this approach is evaluated on the same dataset highlighting the improved value assigned to non-basket-driver items over SIMBA. SIMBAW will provide SMCG retailers with an accurate representation of the total revenue generated by an item.