Research On Service Clustering And Developers' Recommendation Based On Descriptive Information

With the rapid development of 2.0 technology,Service Oriented Architecture(SOA)technology is widely used in service-oriented development mode.This leads to an increasing number of Internet services on a large scale and the service types are also increasingly rich,which makes it difficult to accurately and efficiently find their needs for customers.Therefore,to improve the efficiency of service discovery and to meet the personalized service demand of users have become a difficult task in the field of service discovery.However,the existing service discovery method is only for a certain type of service discoveries and less attention has been attached to RESTful described by natural language,in addition,the current semantic-based service discovery methods require a lot of semantic reasoning calculations,and they are seldom used in practice.Besides,service discovery can help users to find the services they need,but it cannot guarantee that the services are high quality.As we all know,different developers are good at different fields,so recommending the appropriate developers for service development in different areas can ensure the quality of service;but the current number of service developers is so many that it is unrealistic to manually select developers for service development.How to recommend the right developer for the service development has become one of the difficulties in the service development process.But researchers haven't studied this problem both at home and abroad.To solve above problems,this paper aims to study on improvement of service discovery efficiency and management of services against the background of the large scale and various types of service.The current study mainly focuses on the following points:1)Service Clustering Method Based on TF-IDFIn this paper,we propose a service clustering method based on TF-IDF(Referred to WSCBTF-IDF)to achieve service clustering.Firstly,the WSCBTF-IDF method extracts the service function information set from the service description text.Then,based on these service function information sets,the TF-IDF and cosine similarity methods are used to measure the similarity between services.Finally,The k-means algorithm is used to cluster the service set.By clustering of services,this method can reduce the number of users' search services and quickly locate the set of services to meet the needs of users,which can improve the efficiency of service discovery.2)Recommending Developers for Service Building using Naive BayesIn this paper,we propose a new approach based on Naive Bayes(Referred to RDSBNB)to recommend developers for service,which considers the service developer as a kind of service provider recommendation.First,the RDSBNB method uses the service description document to establish a naive Bayesian classifier.Then,based on the classifier,we can classify the new service requirements document(that is,recommend the developer for the service according to requirements document),and get the output of the classifier.Then,the output will be saved to the list,and the list sorted from large to small,and finally,selecting the top k developers to recommend for services.This method not only can help the service management,but also can improve the efficiency of service development.In the end,on the basis of service clustering algorithm based on TF-IDF and Recommending Developers for Service Building using Naive Bayes,Research on Service Clustering and Developers' Recommendation Based on Descriptive Information method has been proposed.Real service data was carried out to verify the algorithm and results showed that this method can not only improve the efficiency of service discovery and management,but also the efficiency of service development.In addition,a prototype system was developed to further demonstrate the feasibility and effectiveness of this method which is of theoretical and practical significance.