Due to WoT (Web of Things) is composed of numbers of resources-limited devices

it is a challenging work that the optimal web service is selected from a similar functionality service group based on the non-functional properties (QoS) without heavy interactive query process in SOA (Service Oriented Architecture).This work focus on the dynamic update process of QoS attributes vector and the optimal service selection algorithm based on the client's feedback in WoT.The service time is estimated by the REM (Random Exponential Marking) which use the concurrent clients number as the price index

and the exponentially weighted average of the current service expense and the previous record.The service availability is estimated by the online probability with the service state detection.The Empirical results based on NS-3.13 simulation show that the accuracy of optimal web service selection based on our presented approaches is significantly better than the simple QoS attributes.