| In order to quickly extract valuable information from massive amounts of data,service recommendation has emerged and has become a research hot topic in recent years.A large amount of existing research work on service recommendation usually uses Quality of Services（Qo S）data as the main basis for recommendation decisions.In reality,Qo S data is often dynamic and changes over time due to geographical location,network conditions and other factors.Currently,some scholars have started to study time-aware service recommendation,but they usually observe the time-varying characteristics of Qo S data as a whole without considering it at a finer granularity,ignoring the intrinsic characteristics of time-series Qo S data.The challenge is to capture the intrinsic characteristics of Qo S data from different time-grained perspectives for service prediction and recommendation.In addition,the data used in the recommendation process is often stored in several different platforms.This requires data sharing between the different platforms,but this process may be subject to the risk of privacy leakage.Therefore,how to make accurate service recommendations based on privacy protection is an important research question.Presently,many scholars have studied the implementation of privacy-preserving service recommendations based on Locality Sensitive Hashing（LSH）techniques.However,the LSH method selects the hash function randomly and has limitations,thus losing some accuracy.Therefore,how to improve the data adaptability of the hashing method and achieve more accurate service recommendation results has become a significant challenge in the current research on privacy-preserving based service recommendation.To address the above challenges,this paper conducts research on time-aware and privacy-aware service recommendation.The details are as follows:（1）The multi-granularity time-aware privacy-preserving service recommendation method Ser Rec EMD-LSHis proposed to address the problem of capturing more fine-grained temporal granularity features.The method is based on the Empirical Mode Decomposition（EMD）method,which improves the original method by treating time-based Qo S data as a basic unit and,capturing different intrinsic characteristics of the time-based Qo S data.Specifically,the Empirical Mode Decomposition-based approach divides the sequence of service quality changes over time into features with different time granularity,and then finds similar time sequences based on LSH techniques with privacy protection,in order to predict the rating values of services that users may be interested in,and recommends the top-N services with the highest ratings to users.Finally,a case study is provided to demonstrate the feasibility of the proposed Ser Rec EMD-LSHapproach.（2）The Iterative Quantization-based time-aware privacy-preserving service recommendation method Ser Rec ITQ-Timeis proposed to address the enhancement problem of the hashing process in privacy-preserving.the method mainly uses the Iterative Quantization（ITQ）method,which divides the Qo S data of time by a data-trained hash function into data less hash codes for recommendation.Specifically,the LSH method for privacy preservation in time awareness is improved using Iterative Quantization approach,focusing on the perspective of training the hashing process to improve the fitness of the hash to the data.The privacy-preserving hashing process is then trained using an unsupervised approach for the optimisation problem of binary code nearest-neighbour search in time-aware scenarios,aiming to improve the accuracy and applicability of the privacy-preserving process to the data classified process.Finally the effectiveness of the method is validated on a real dataset,WSDream,and the experimental results demonstrate that the Ser Rec ITQ-Timemethod is more accurate than other comparative methods. |
