Domain Knowledge Based Service Clustering And Personalized Recommendation Method

With the rapid growth of the number of Internet software,a large number of Web services with similar functions are produced.The difficulty of service discovery has been caused by the addition of new users.Therefore,it is very important to propose a new service clustering and recommendation algorithm as well as domain knowledge discovery method.As with the traditional service recommendation method system,it is first necessary to enrich the user's application domain knowledge of related services,in order to increase user input query auxiliary words,and solve the problem of cold start.Therefore,there are great significance to service discovery and recommendation by domain knowledge utilization and on-demand extraction.Domain knowledge in this paper refers to the general knowledge of services summarized by domain experts.Especially,the construction of knowledge networks and the discovery and utilization of hotspot knowledge will contribute to the development of software requirements and service recommendations.Research on appropriate knowledge network construction technology and hotspot knowledge discovery technology can not only make up for the service recommendation difficulties caused by the lack of user domain knowledge,but also support the existing knowledge network evolution needs.In addition,it is a very tedious work by processing disordered and a large number of unprocessed text datasets in PWeb.Orderly organization of service knowledge from the perspective of complex network model structure can improve the accuracy of service discovery and recommendation.Therefore,knowledge-oriented discovery and organization can provide important theoretical and practical significance for Internet service precision recommendation.The main contents of this paper include the following 3 innovations:(1)How to build domain knowledge of service based on evolutionary network in the absence of user domain knowledge.In this study,we propose an adaptive growth model for knowledge evolution networks.This research is based on the Internet services and PWeb website services.It is a knowledge evolution network that studies how to obtain service knowledge from natural language description knowledge and merge similar knowledge nodes.This paper studies how to predict the development of service knowledge by calculating the degree of hotspot of knowledge nodes through the degree of input and output,and provides a quantitative basis for service recommendation.(2)How to use similar words to enhance LDA clustering model to find similar services in service documents described by natural languages.Aiming at this problem,a cosine similarity method based on Word Embedding is proposed to find similar words of keywords to assist LDA to cluster services in various fields,so as to improve the accuracy of service clustering.Aiming at the problem that the original auxiliary vocabulary dictionary implies a large number of noise words,how to effectively filter the vocabulary beneficial to improve the clustering accuracy rate,This paper proposes a method based on Word2 vec three-layer neural network combined with TF-IDF method to generate the characteristic threshold to eliminate noise words.And it can automatically find the threshold through the hierarchical clustering,we found that the high-quality auxiliary vocabulary after removing the noise words can effectively improve the service clustering purity by clustering result of the 10 domain of the real PWeb website.(3)How to find personalized services in line with massive and disorderly PWeb services,and recommend personalized services to users.In this study,we proposed a method of mining and recommending Web services based on RGPS Metamodel.This method modeling by roles,goals,processes,and services be directed against specific domain issues(SDI).We defined RGPS elements relationship network basis on the existing RGPS demand meta-model.The related operations in each model are model by temporal logic of actions(TLA),and use a logical sequence to achieve complex element organization.This paper designs a domain-specific and RGPS on-demand service organization method,and organizes and manages related services from different perspectives and levels.According to different user needs,we design a service recommendation algorithm is designed,which enables users to quickly and accurately find the best service to meet their needs and related service organizations.Based on the above three innovations,we developed corresponding experimental demonstration tools,which consists of a domain knowledge-based service acquisition and clustering analysis tool,and an evolutionary network generation tool.We selected the service dataset crawled from ProgrammableWeb for experiments.Experiments show that the fusion of knowledge evolution network through RGPS metamodel can effectively find services for users.