QoS Trust Rate Prediction for Web Services Using PSO-Based Neural Network

Trustworthiness is an important indicator for service selection and recommendation in the environment of cloud computing. However, how to predict or evaluate the trust rate based on the multifaceted QoSs is not an easy task. According to the existing studies, intelligent technique is a rational way to settle this problem. Due to complicated and non-linear relations between service's QoS values and its final trust rate, neural network has been validated as an effective approach for trust rate prediction. Unfortunately, the parameter setting of neural network plays obvious impact on its prediction performance. In the paper, particle swarm optimization (PSO) is introduced to enhance neural network by optimizing its initial settings. In the hybrid prediction algorithm named PSONN, PSO is used to search the appropriate parameters for neural network to conduct the further prediction. In order to investigate the effectiveness of PSO-NN, the experimental analysis is performed on public QoS data set. The results show that our proposed approach has better performance than the basic classification methods, and significantly outperforms the traditional neural network (BPNN) in terms of prediction precision.