An active database architecture for knowledge-based incremental abstraction of complex concepts from continuously arriving time-oriented raw data

摘要

An effective solution to the tasks of continuous monitoring and aggregation querying of complex domain-meaningful concepts and patterns in environments featuring large continuously changing data sets is very important for many domains. Typical domains include: making financial decisions, integrating intelligence information from multiple sources, evaluating the effects of traffic controllers’ actions, detection of security threats in communication networks, planning and monitoring in robotics, and management of chronic patients in medical domains. In this paper, we present a general domain-independent method for an effective solution of these two tasks. Our method involves incremental creation of meaningful, interval-based abstractions, from raw, time-stamped data continuously arriving from multiple sources, which is supported by the accumulation and continuous validation of the created abstractions. We implemented our method in the Momentum system, which is an active knowledge-based time-oriented database—a temporal extension of the active-database concept that we propose for incremental application of knowledge to continuously arriving time-oriented data. We evaluated the Momentum system in a medical domain within a database of 1,000 patients monitored after bone-marrow transplantation, and a knowledge base of complex abstractions regarding more than 100 raw-data types and about 400 concept types derivable from them. Initial evaluations are highly encouraging with regards to the feasibility of the whole approach.