Knowledge Discovery in Databases to Support Business Decision in Customer Relationship Management

Student Name: Manepalli P.K.V.Subba Rao

This was a comprehensive case study in data mining from data warehouses containing data on demographics and purchase patterns of``members'' (card holders) of a non-grocery business with nation-wide chain of retail outlets for better customer relationship management. In particular, the main business objectives were Customer Segmentation, Customer Profiling and Product Affinity analysis for generating ideas for cross selling and general understanding of customer preferences. Customer segmentation was done with respect to customer demographics, purchase patterns (after defining appropriate stochastic processes) and purchase interests using an artificial neural network with unsupervised learning involving Kohonen's feature map. Customer Profiling was done using a feed-forward neural network with customer demographics as the inputs and four ``purchase pattern'' segments, developed in the previous step, as the output. Product Affinity analysis was performed by computing various joint, conditional and ratios of probabilities (which are respectively called the support, confidence and lift in data mining jargon) of sale of various products and brands and some interesting ``rules'' regarding product$\rightarrow$product, product$\rightarrow$brand, brand$\rightarrow$product and brand$\rightarrow$ brand were discovered. These discovered Product Affinity rules and the earlier Customer Segmentation and Customer Profiling results were then leveraged to generate Cross Selling ideas, apart from providing some specific recommendations at strategic, tactical and operational level.