Features

The tenth annual ACM SIGKDD conference (KDD 2004), the premier international conference on knowledge discovery and data mining, was held in Seattle August 22 - 25. The conference succeeded again to provide a forum for academic researchers and industry and government innovators to share in their advances in the identification of valid, novel, potentially useful, and ultimately understandable patterns in data.

Spam filtering is emerging as one of the major application areas, with a whole tutorial and a number of papers devoted to it.

A panel on Data Mining and Privacy explored the implications to data mining in the wake of Congress closure of Total Information Awareness. Computer experts presented ways to perform privacy-preserving data mining (another hot new topic), while a law professor from Berkeley recommended ways to create a legal framework for data mining within federal government.

Summary of panels and KDD Cup presentations will appear in forthcoming issue of SIGKDD Exploration. (www.acm.org/sigkdd/explorations).

The conference was also videotaped for the first time. Details on getting the DVDs will appear in KDnuggets and on www.acm.org/sigkdd/kdd2004/ in a couple of weeks.

A highlight of the conference was the announcement of the KDD Cup competition results (http://kodiak.cs.cornell.edu/kddcup/).

This year's competition featured tasks in particle physics and bioinformatics. The goal of the particle physics task was to learn a classification rule that differentiates between two types of particles generated in high energy collider experiments. The goal of the bioinformatics task was to predict which proteins are homologous to a native sequence. A variety of performance criteria were used, such as accuracy, squared error, cross entropy, and ROC area.

Eight workshops were held in conjunction with the conference. Along with data mining standards, topics addressed data sources such as bioinformatics, temporal and sequential data, multi-relational data, multimedia data, link data, and web and semantic web data.

KDD-2004 was sponsored by six gold sponsors and one bronze sponsor: Amazon, Boeing, Fair Isaac, Microsoft, SAS, and SPSS were gold sponsors; Mitre was a bronze sponsor. AAAI, IBM, SIGMOD, and Elder Research contributed to student awards. The Kluwer Machine Learning Journal sponsored the best student paper award. Over a dozen companies had exhibit booths.

Next year KDD-2005 conference will take place in Chicago, IL. from August 21 through 24.