A METAPATTERN-BASED AUTOMATED DISCOVERY LOOP FOR INTEGRATED DATA MINING - UNSUPERVISED LEARNING OF RELATIONAL PATTERNS

Metapattern (also known as metaquery) is a new approach for integrated data mining systems. Different from a typical ''tool-box'' like integ ration, where components must be picked and chosen by users without mu ch help, metapatterns provide a common representation for intercompone nt communication as well as a human interface for hypothesis developme nt and search control. One weakness of this approach, however, is that the task of generating fruitful metapatterns is still a heavy burden for human users. In this paper, we describe a metapattern generator an d an integrated discovery loop that can automatically generate metapat terns. Experiments in both artificial and real-world databases have sh own that this new system goes beyond the existing machine learning tec hnologies, and can discover relational patterns without requiring huma ns to prelabel the data as positive or negative examples for some give n target concepts. With this technology, future data mining systems co uld discover high-quality, human comprehensible knowledge in a much mo re efficient and focused manner, and data mining could be managed easi ly by both expert and less expert users.