Evaluation of first order Bayesian networks for context modeling and reasoning

There are numerous implementations of context aware application frameworks. Knowledge is being inferred in diverse ways across numerous platforms. A disappointing observation is the limited reuse of context reasoning and knowledge. From the perspective of a new context aware application developer, existing frameworks may not be sufficient for the following reasons: i) insufficient knowledge representation formalisms; ii) limited generalized APIs for reasoning; and iii) inefficient and inflexible reasoning choices that fail to meet application needs. Augmenting existing frameworks with a generalized hierarchical hybrid context reasoning engine (HyCoRE) could improve their reusability. However, a middleware that offers a sufficient variety of optimizable techniques for reasoning across heterogeneous contexts is not openly available. As a step towards HyCoRE, we evaluate MEBN, a first order Bayesian network knowledge formalism, for suitability in data modeling and reasoning for pervasive computing contexts.