Research And Implementation Of Web Service Recommendation Model Based On QoS-Aware Of Functional And Geographical Features Data

The development of Internet technologies and services has greatly expanded the breadth and depth of people's production and life.It is difficult for users to find suitable services in massive Web services,making personalized recommendation of Web services one of the most important challenges in the field of service computing.At present,many researches and technologies generate recommendations based on client QoS,but lack of consideration of the impact of server function attributes and user geographic location on recommendation results in the sparseness of service call matrix data.For this problem,this paper does the following work:Firstly,a Web service functional similarity calculation model based on text information mining is constructed,and service clustering is based on the model.This paper uses the URL in the dataset to obtain the WSDL document of the service,and extracts the context feature to obtain its functional description and normalize it.Then build the similarity calculation model based on context information,and use K-means++ algorithm to perform similar service clustering.Secondly,this paper conducts user clustering based on hierarchical geographic neighborhood and modified cosine similarity calculation model to obtain a set of users with similar service preferences.The impact of geographical location on users' Web service preferences is analyzed.The QoS preferences of users and their neighbors are positively correlated,which makes it necessary to consider geographical neighbors in the recommendation model.We use the modified cosine similarity calculation formula as a tool to calculate the degree of difference between users,and design a user clustering algorithm based on hierarchical geographic neighborhood.A data set is processed experimentally to obtain a similar set of users.Finally,a matrix decomposition model based on similar services and user clustering is proposed to predict QoS,and a tag model is used to generate personalized service recommendations.In order to alleviate the inaccurate impact of recommendation results caused by data sparseness,we propose a matrix decomposition model(RMUSC)to perceive QoS,and design a gradient descent algorithm to converge to determine the predicted QoS value;we introduce a label model to make predictions The QoS is more efficient in producing recommendation results.The main principle is to use the triplet relationship of <service,label,QoS> for label modeling.Simplifies the relationship between users and services.Finally,our model is analyzed by experiments to have better recommendation performance.