An autonomic problem determination system using adaptive multi-levels dictionaries and singular value decomposition techniques

Abstract

An autonomic Problem Determination (PD) system can adapt to changing environments, react to existing or new error conditions and predict possible problems from a collection of seemingly unrelated events and take actions proactively. In this paper, we propose such a system with the use of the Singular Value Decomposition (SVD) technique and dynamic and adaptive multi-levels dictionaries. Our proposed PD system applies an iterative method that enables a dynamic interaction between a set of sparsely related events and the current dictionaries, with its entries being continuously updated to reflect the relative importance of each event. Updating the dictionaries triggers an update of the SVD matrix, thereby accelerating the convergence of the SVD matrix. Our PD system can capture knowledge in a hierarchical form for complex knowledge representation. It does not require a formal knowledge model or intensive training by examples. It is an efficient system with sufficient accuracy for autonomic and adaptive problem determination.