Interestingness of association rules in data mining: Issues relevant to e-commerce

The ubiquitous low-cost connectivity synonymous with the internet has changed the competitive business environment by dissolving traditional sources of competitive advantage based on size, location and the like. In this level playing field, firms are forced to compete on the basis of knowledge. Data mining tools and techniques provide e-commerce applications with novel and significant knowledge. This knowledge can be leveraged to gain competitive advantage. However, the automated nature of data mining algorithms may result in a glut of patterns — the sheer numbers of which contribute to incomprehensibility. Importance of automated methods that address this immensity problem, particularly with respect to practical application of data mining results, cannot be overstated. We first examine different approaches to address this problem citing their applicability to e-commerce whenever appropriate. We then provide a detailed survey of one important approach, namely interestingness measure, and discuss its relevance in e-commerce applications such as personalization in recommender systems. Study of current literature brings out important issues that reveal many promising avenues for future research. We conclude by reiterating the importance of post-processing methods in data mining for effective and efficient deployment of e-commerce solutions.