In order to handle the uncertainty of data used in quality of service (QoS) prediction, increase the confidence value of prediction results, and make the QoS prediction more reliable, a probability model for quantifying the confidence in QoS prediction was built. In the process of building the model, the number of QoS data items used in prediction, the data fluctuation (data deviation), and the data decay over time were considered. The results show that the number of data reflects the impact of the number of QoS data used in prediction on the reliability of the prediction result, the data deviation reflects the consistency degree of the actual value and predicted value of QoS in service, and the data decay reflects the impact of data on the prediction result over time. The simulated test indicates that the confidence model can help consumers effectively select services based on their requirements.