Web service recommendation using memory-based and model-based collaborative filtering algorithms

With the ever-growing volume of web services on the internet, recommendation systems are playing a very important role to overcome such information burden. In this study, we recommend applicable services quickly and accurately using Collaborative Filtering (CF) techniques. The novel clustering approach by ontology generation that is based on domain specificity and service similarity is used to solve cold-start and data sparsity issues. After solving the issues, the approach is continued for the recommendation based on memory-based and model-based CF algorithms. The memory-based method calculates the similarity between users and then the trust value between users were calculated to predict the user ratings. The model-based method is applied for each separate cluster by the K-Nearest Neighbour (KNN) algorithm and user ratings were predicted by the Singular Value Decomposition (SVD). Among the proposed two approaches the model-based approach performed with higher accuracy than the memory-based approach when compared with existing approaches.