A similarity-enhanced hybrid group recommendation approach in cloud manufacturing systems

With the development of cloud manufacturing (CMfg), a huge amount of services appears on the Internet, which makes recommender systems in CMfg service a promising research field. To this end, recent studies mainly focus on solving individual recommendation to meet the requirements of every user. However, due to the time complexity problem, ‘Many-to-Many’ recommendation mode is increasing in real applications. To implement such a group recommendation is very challenging, because the system not only needs to achieve high recommendation quality but also satisfies user clusters in an even way. Therefore, we propose a similarity-enhanced hybrid group recommendation approach named HGRA for cloud manufacturing. Specifically, we implement the HGRA system by three main components. Firstly, an enhanced user similarity measuring approach is designed to identify a similar user group based on the non-linear model Proximity-Significance-Singularity (PSS) and Kullback-Leibler (KL) distance algorithms. Secondly, a set of user subgroups are further clustered using K-medoids algorithm, in which additional information similarity is calculated by making full use of functional information about the users. Thirdly, a weighted ranking aggregation model is established to generate a recommendation list according to the representative user of each subgroup. The performance of our system is tested by two data sets from real-world cloud manufacturing systems. The experimental results demonstrate the feasible and effectiveness of our approach, compared with some state-of-the-art benchmark solutions, especially in CMfg systems.