Interactive and Online Data Mining

Cornell University, Ithaca NY

Investigators

Abstract

Data mining is one of the very promising information technologies today. We are concentrating our research into two new topics in data mining. First, instead of optimizing construction of a complete data mining model, parts of the model are incrementally constructed as guided by user feedback in order to reduce the long construction times of data mining models. The methods enable interactive response times, and this research is expected to result in a paradigm shift away from batch-oriented mining to interactive data mining. Second, data mining has traditionally been performed over static datasets, and mining algorithms could afford to read the input data several times. This traditional approach is referred to as offline data mining. This research addresses the online mining of high-speed data streams. The goal is to develop data mining systems that never stop working, incorporate new records immediately as they arrive, and update current and construct new models continuously. This new approach is referred to as online data mining. Anticipated applications include the discovery of patterns in network data, click-stream data, text data, and/or biological databases.