Research On Service Recommendation Algorithm Based On Collaborative Filtering Integrating Semantic Similarity

In the Web 2.0 era,internet-based Web services are increasingly favored by developers due to their high efficiency,low cost,and loose coupling features,and Mashup has attracted extensive attention from researchers as a software that combines multiple single-function Web services to provide complex application scenarios.However,it is difficult for developers to manually select Web services to meet the increasingly complex user needs in the development process of Mashup.How to find the most suitable service set from various available Web services is a challenging research subject.Recently,collaborative filtering algorithm based on matrix decomposition has been widely used in the service recommendation model,but it cannot effectively capture the complex interaction between them in the sparse Mashup-Service invocation matrix,which will lead to low recommendation performance.Towards this problems we proposed a collaborative filtering recommender algorithm method integrating content similarity.In addition,social-based recommender algorithm is also applied in Web service recommendation tasks,but it could not effectively facilitate auxiliary info,and we noticed that another shortcoming of the collaborative filtering-based recommendation algorithm on is that two web service recommendation scenarios are not considered.Towards this problem,we proposed a collaborative filtering Web service recommendation algorithm integrating HIN and topic models.The main contributions of our work are as follows:1.We proposed a collaborative filtering service recommendation algorithm integrating semantic similarity,the CF part was used to captured the non-linear relationship between Mashup-Service,and the content part was used to capture the semantic similarity between them,and the algorithm integrates the collaborative filtering module and the content similarity extraction module into the deep neural network seamlessly to predict the rating of Web services,and finally accomplished the recommender tasks.2.We proposed a collaborative filtering Web service recommendation approaches integrating NIN and topic model.The offline module of the algorithm preprocesses the Mashup-Service information to reduce the search space when the online service is recommended by using the topic model,and it calculates the similarity between Mashups through the heterogeneous information network;the online module predicted the rating of Web Service according to the existing Mashup-service rating information based on the collaborative filtering technology,and the corresponding Web service recommendation list is generated according to different service recommendation scenarios,and finally the recommender task is completed.A set of experiments are conducted for these two Web service recommendation approaches with dataset from the ProgrammableWeb website.The result shows that our algorithm has achieve significant performance improvement on various evaluation metrics comparing with the state-of-the-art Web service recommendation approaches,and could effectively perform Web service recommendation.