Design of the Autonomous Fault Manager for learning and estimating home network faults

This paper proposes a design of software autonomous fault manager (AFM) for learning and estimating faults generated in home network. Most of the existing researches employ rule-based fault processing mechanism, but those works depend on the static characteristics of rules for a specific home environment. Therefore, we focus on a fault estimating and learning mechanism that autonomously produces a fault diagnosis rule and predicts an expected fault pattern in the mutually different home environment. For this, the proposed AFM extracts the home network information with a set of training data using the 5W1H (Who, What, When, Where, Why, How) based contexts to autonomously produce a new fault diagnosis rule. The fault pattern with high correlations can then be predicted for the current home network operation pattern.