A Constrained Learning Approach to the Prediction of Reliability Ranking for WSN Services

Wireless Sensor Network Service Applications (WSAs) are playing an important role in Wireless Sensor Network (WSN), which bridge the gap between WSN and existing widely deployed Service-Oriented Architecture (SOA) technologies. Function properties of WSN services are important, which assure correct functionality of WSA. Meanwhile, nonfunctional properties such as reliability might significantly influence the client-perceived quality of WSA. Thus, building high-reliability WSA is a critical research problem. Reliability rankings provide valuable information for making optimal WSN service selection from functionally equivalent service candidates. There existed several methods that can conduct reliability ranking prediction of WSN services. However, it is difficult to evaluate which one is better than another, because those acquire different rankings with different preference functions. This paper proposes a constrained learning prediction of reliability ranking approach for WSN services on past service usage experiences of other WSAs, which can achieve higher accuracy and improve the performance by pruning candidate services. To validate the authors' approach, large-scale experiments are conducted based on a real-world WSN service dataset. The results show that their proposed approach achieves higher prediction accuracy than other approaches.