Research On Construction Of Reliable Trust Network In Distributed Environment

Malicious buyers in the e-commerce platform will perform malicious attacks by providing false evaluations and other methods to influence the shopping choices of other buyers.Although the traditional serial trust model can effectively resist malicious attacks,however,in the face of large-scale data,its computing efficiency is low,and it cannot meet the computing needs of massive data.Therefore,it is of great significance to study the parallel design of trust models and build a reliable e-commerce platform trust network.This article aims to improve the computational efficiency of the Multi-agent Evolutionary Trust Model(MET)and Trust-aware Social Influencer Finding(TrustINF)algorithms.The main work includes the following aspects:(1)Aiming at the problem that the evaluation data in the data set is too sparse and the MET fitness value cannot be calculated,a filling algorithm is introduced in combination with the trust relationship to predictively fill the evaluation data,which improves the accuracy of the filling result.Aiming at the problem of low timeliness of MET in big data environment,a parallel MET model based on Spark—SparkMET is proposed.This model adopts a master-slave architecture,where the master node performs mutation,crossover,selection processes,and the slave nodes calculate the fitness value.Aiming at the problem of data skew in the parallelization process of MET,a new data partitioning strategy—LBP algorithm is proposed.This algorithm re-partitions the input data according to the appropriate partition label before SparkMET calculates the fitness value,which can further improve the computing efficiency of MET.Finally,experiments are performed on the Epinions dataset,the experimental results show that the parallel model proposed in this paper can greatly improve the computing efficiency of the original model.(2)Aiming at the problem that the TrustINF algorithm has a low recognition rate for malicious buyers,the TrustINF algorithm was improved by introducing a new slope calculation method and a trust number change rate factor.Then,by adding a suspend function to simulate large-scale data,on this basis,the parallelization of the improved TrustINF algorithm on the Spark platform is realized.Finally,experiments are performed on the Epinions dataset,the experimental results show that the improved algorithm not only improves the computing efficiency significantly,but also has a malicious buyers recognition rate of 71%,which can effectively resist malicious attacks and build a reliable trust network.