Mining retail outlet transaction data
Abstract
Retailing is increasingly becoming a high performance sector in the Kenya economy and retailers are fast seeking a competitive edge through-technology. We describe the exploitation of Data Mining techniques and in particular association rule mining to analyze various baskets of a popular retail shop in Kisumu. The aim of the basket analysis was to allow retailers to quickly and easily look at the size, content and value of their customers' products to understand patterns, affinities and associations with a view to identifying cross-sell opportunities, improve shop floor layout and organization, encourage impulse buying driving promotions and advertisement based on the database intelligence and identify new business opportunities.
We used CRISPDM (Cross-Industry Standard Process for Data Mining) methodology for data mining and predictive analytics. CRISP DM was adopted because of its ability to iteratively move back and forth in all the six stages of data miming namely business understanding, data understanding, data preparation, modeling, evaluation and deployment.
The dataset used for this case study contained data for both loyal and non loyal customers. We cleaned the data and converted it to binary format for analysis. The data was divided into two partitions in equal portions of two months periods. The results observed were that regularities in both partitions were fairly consistent such as the rules and itemset generated. Best transacted items revolved around basic Fast Moving Consumer Goods. With the regularities observed a floor plan and a stimulus response model were proposed for the retail shop with a view to improving impulse buying therefore improving sales. Key words: Data mining, Association rule mining and market basket analysis.
Publisher
University of Nairobi, Kenya