Industry players agree that data mining technology, as one of the strong pillars in CRM (Customer Relationship Management), plays an important role in discovering customer behavior. However, applying the technology in business applications requires mining experts. Without expertise in adjusting more than 20 parameters in the mining process, data mining means nothing valuable to enterprises. We introduce this R&D project to propose a fundamental change to traditional data mining tools to minimize the involvement of technical expertise in adjusting parameters.

The first problem we tackle is the mining of association rule which has found applications in many different business arenas. We propose to consider a more practical problem, roughly speaking, it is to mine the top N most frequent itemsets, new algorithms are introduced for mining N-most interesting itemsets. Another application of association rule is in the inventory control problem, where the products carried by a retailer or a service provider are varied from time to time to optimize the profits. For such a problem the effect of cross-selling should be considered. We strongly believe the research output would benefit the Hong Kong-based enterprises by the way of understanding their customer behavior through automated and manageable data mining platform.