Efficient context-aware selection based on user feedback

Adaptive services in pervasive environments are based on the correct detection of context. However, sensor malfunctions and inappropriate inference in regards to dynamic environments can lead to incorrect context detection that is unintended by the user. In addition, an appropriate context-aware method needs to be accurate even when environmental conditions change. Feedback from the user is one of the methods used to correctly acquire contextual information and feedback data that can be used to select the adaptive context-aware method. This paper presents a scheme that evaluates context-aware methods based on the feedback data from the user. The evaluation is performed by comparing the feedback data within the context of the currently running context-aware methods. The error rates of all the context-aware methods are calculated and the service provider then selects the appropriate context-aware method which possesses the smallest error rate amongst them. Experiment results show that the proposed method improves the context-aware rate by up to 10.3% compared to the trust-worthiness based method and 16.4% compared to the voting method.