Towards a Trust Prediction Framework for Cloud Services Based on PSO-Driven Neural Network

Abstract

Trustworthiness is an important indicator for service selection and recommendation in the cloud environment. However, predicting the trust rate of a cloud service based on its multifaceted quality of services (QoSs) is not an easy task due to the complicated and non-linear relations between service's QoS values and the final trust rate of the service. According to the existing studies, the adoption of intelligent technique is a rational way to attack this problem. Neural network (NN) has been validated as an effective way to predict the trust rate of the service. However, the parameter setting of NN, which plays an important role in its prediction performance, has not been properly addressed yet. In the paper, particle swarm optimization (PSO) is introduced to enhance NN by optimizing its initial settings. In the proposed hybrid prediction algorithm named PSO-NN, PSO is used to search the appropriate parameters for NN so as to realize accurate trust prediction of cloud services. In order to investigate the effectiveness of PSO-NN, extensive experiments are performed based on public QoS data set, as well as in-depth comparison analysis. The results show that our proposed approach has better performance than basic classification methods in most cases, and significantly outperforms the basic NN in the terms of prediction precision. In addition, PSO-NN demonstrates better stability than the basic NN.

Keywords

• Cloud services
• quality of service (QoS)
• trust prediction
• neural network
• particle swarm optimization (PSO)