A DATA MINING FRAMEWORK FOR PRODUCT BUNDLE DESIGN AND PRICING

Abstract

Product bundling is a marketing strategy that has been widely studied in research literature and extensively used in practice. With the growing quantity of products and huge possible bundling combinations, it is necessary to develop algorithmic approaches to determine which items should be in a profitable bundle, which bundling strategy is most profitable, and what the proper price is for a bundle. Previous studies have put forward many approaches for bundle design, but they have many limitations. Consumer’ behaviors may be not in accordance with their statements in a survey, thus the transaction data is a more reliable source to predict their purchase behaviors. As consumers’ demand and market supply will fluctuate continuously, fail to consider price elasticity of demand (PED) will cause biases for prediction, where PED is used to measure consumers’ abilities and willingness to pay for certain products. In this thesis, we propose a data mining framework which incorporates the time value of money in data mining tasks, and it is capable of determining the product combination and price of a bundle in order to maximize the revenue. We also apply association mining to generate meaningful candidate bundles and reduce computation cost. This framework analyzes consumer and product data, taking demand and inflation factors into consideration, to fill in the gaps as mentioned. We also demonstrate the efficiency of this data mining framework through experiments and simulations.