Market Basket Analysis with Statistically Improved Association Rules Considering Product Details

OpenAlex · English

Abstract E-commerce companies need to direct customers to choose the wished products. A significant challenge in the e-commerce domain forces to use of technology and data-oriented solutions. So, investigating the sales transactions is necessary thanks to the high number of product varieties for product recommendation. Although creating association rules with many algorithms is common, this study improves the traditional association rule mining (ARM) method by adding a different perspective, named statistical settings, to create more insights for managerial implications. After generating rules, it checks the statistical significance of the correlation, considering some product-specific details, such as product name, discount rates, and the number of favorites. The study concluded with exciting insights. For example, the company should recommend products with a lower discount rate if the customer puts products in the basket with a high discount rate. It also enriches traditional rules by adding product features. When the total number of favorites is less than 7500, and the discount rate is less than 75\%, the similarity ratio of the Conclusion products should be less than 0.50.