Time-Aware Service Recommendation for Mashup Creation

Abstract

Web service recommendation has become a critical problem as services become increasingly prevalent on the Internet. Some existing methods focus on content matching techniques, while others are based on QoS measurement. However, service ecosystem is evolving over time with services publishing, prospering and perishing. Few existing methods consider or exploit the evolution of service ecosystem on service recommendation. This paper employs a probabilistic approach to predict the popularity of services to enhance the recommendation performance. A method is presented that extracts service evolution patterns by exploiting latent dirichlet allocation (LDA) and time series prediction. A time-aware service recommendation framework is established for mashup creation that conducts joint analysis of temporal information, content description and historical mashup-service usage in an evolving service ecosystem. Experiments on a real-world service repository, ProgrammableWeb.com, show that the proposed approach leads to a higher precision than traditional collaborative filtering and content matching methods, by taking into account temporal information.

Temporal‐aware and sparsity‐tolerant hybrid collaborative recommendation method with privacy preservation

“…There have been some researches discussing temporal influence to recommendation accuracy, but most of them focus on discussing the effect to user preferences or popularity. 12,[39][40][41] Chen et al 12 and adapt factorization techniques to learn user-group affinity based on two different implicit engagement metrics. However, there are but few works considering the temporal influence to QoS performance of cloud services.…”

“…Thus, temporal information has great impact to the prediction performance of recommender systems. There have been some researches discussing temporal influence to recommendation accuracy, but most of them focus on discussing the effect to user preferences or popularity . Chen et al propose a time‐sensitive personalized pairwise ranking model to address the repeat consumption problem based on the behavioral features extracted from user implicit feedback.…”

“…There have been some researches discussing temporal influence to recommendation accuracy, but most of them focus on discussing the effect to user preferences or popularity. 12,[39][40][41] Chen et al 12 propose a time-sensitive personalized pairwise ranking model to address the repeat consumption problem based on the behavioral features extracted from user implicit feedback. Wang et al 40 implement a time-varying factorization to capture the temporally extended nature of group engagement and adapt factorization techniques to learn user-group affinity based on two different implicit engagement metrics.…”

Lazy collaborative filtering with dynamic neighborhoods

“…In recent years, much research work has been done in both academia and industry on developing new approaches for dealing with scalability and providing an effective recommendation based on user’s requirements. In Zhong et al ’s 2015 study, a time-aware service recommendation model is developed for mash-up creation using temporal information. Latent Dirichlet allocation and time predictions are exploited to enhance the performance of the RS.…”

“…In such situations, time-cost (Chen and Zhou, 2015) is used, in which time is discussed in terms of cost and incorporated into the recommender methods. The third approach is the time-decay (Zhong et al , 2015; Hariri et al , 2015; Stefanidis et al , 2013; Noam et al , 2011) approach which has been widely experimented over the couple of years. The user preference constantly evolves with time, resulting in lesser impact of the older ratings.…”