Grand Challenges on the Road to Practical Data Mining Systems

by Usama Fayyad, DMX Group, USA

The past two decades has seen a huge wave of computational systems for the "digitization" of business operations from ERP, to manufacturing, to systems for customer interactions. These systems increased the throughput and efficiency of conducting "transactions" and resulted in an unprecedented build-up of data captured from these systems. The paradoxical reality that most organizations face today is that they have more data about every aspect of their operations and customers, yet they find themselves with an ever diminishing understanding of either. Data Mining has received much attention as a technology that can possibly bridge the gap between data and knowledge. While some interesting progress has been achieved over the past few years, especially when it comes to techniques and scalable algorithms, very few organizations have managed to benefit from the technology. Despite the recent advances, some major hurdles exist on the road to the needed evolution. Furthermore, most technical research work does not appear to be directed at these challenges, nor does it appear to be aware of their nature. This talk will cover these challenges and present them in both the technical and the business context. The exposition will cover deep technical research questions, practical application considerations, and social/economic considerations. The talk will draw on illustrative examples from scientific data analysis, commercial applications of data mining in understanding customer interaction data, and considerations of coupling data mining technology within database management systems. Of particular interest is the business challenges of how to make the technology really work in practice. There are many unsolved deep technical research problems in this field and we conclude by covering a sampling of these.

This page has been accessed times since August 28, 2003.