Research On Time-aware Service Recommendation Method

Service recommendation is a hot topic in service computing.With the maturity of service computing technology and the rapid development of the service market,users’ demand for services increase significantly,service recommendation plays an important role gradually in people’s daily life.People enjoy the convenience brought by services,meanwhile,service recommendation faces the problem of how to accurately recommend services to users.With the popularity of mobile devices and the Internet,users’ search for services is real-time,so time-aware series recommendation become the hot issue in recommendation research.Time facors affect the function,status of the serviceand user preferences,and challenges the effectiveness of the recommendation algorithm.This paper studies the research status at domestic and foreign,and launches the following innovative work:(1)Aiming at the problem that user preferences change with time,a collaborative filtering method for time-varying preferences is proposed.Taking the short-term preference,which is most susceptible to time factors,as the research core,the short-term preference is obtained by establishing a preference analysis model,and a scoring reference model is constructed to obtain the change of users’ short-term preference and the current service reference score.The time decay function is improved to make it more practical.Finally,the similarity between users is obtained by collaborative filtering method,and the service rating of users is predicted,so as to make real-time service recommendation.The proposed method is verified by real data sets,and the experimental results show that this method is superior in prediction performance.(2)Aiming at the problem that time series prediction is influenced by time factors,multi-feature time-aware series service recommendation method is proposed.Take time factor and sequence characteristics as the research core,time series is decomposed to obtain several stable multi-feature subsequences,multi-feature input patterns and time stamp series are constructed as the input of Gate Recurrent Unit(GRU).Then,the sequence is predicted according to the length of future prediction time step,and real-time service recommendation is made.The proposed method is compared with other methods.The real experimental data shows that this method can successfully obtain multi-feature subsequences and improve the accuracy of service recommendation under time-awares.