Software Service Recommendation Method Based On User Context Awareness

With the rapid development of a new generation information technology,such as cloud computing,big data,Io T,AI and mobile internet,and the rapid popularization of the concept of "everything is a service",smartphones are becoming more and more important in people’s lives.Correspondingly,the number of software services（ie,mobile application software）in the mobile application market is also continuing to grow.When selecting from such a huge set of candidate software services,users often can not easily find the appropriate software services.Therefore,how to efficiently recommend suitable software services for users has become a hot issue in the field of recommendation.In order to meet the needs of users to find suitable software services conveniently,many scholars have explored the problem of service recommendation.However,the current classic recommendation methods generally have problems such as unsatisfactory accuracy of recommendation results,incomplete consideration of user personalized characteristics,and difficulty in mitigating cold start problems.Therefore,from the perspective of mobile application market operators,this thesis fully perceiving the context information of users in the process of recommending software services,and then effectively realizing personalized recommendation of software services.This thesis mainly includes the following contents:（1）By collecting the user’s basic information,evaluation information,download information and other contexts,the user is depicted from the three dimensions of basic attributes,behavioral attributes and psychological attributes.Among them,DT-Kmeans algorithm is used for label clustering for long text comment information.Then,the user portrait similarity is calculated and combined with user score similarity,and a collaborative filtering algorithm is proposed.Compared with the classical collaborative filtering algorithm,the proposed algorithm has better accuracy and can better alleviate the cold start problem.（2）At present,most researchers apply general knowledge graph or domain knowledge graph to service recommendation field,but this results in insufficient consideration of user’s personalized characteristics.In order to solve this problem,this thesis firstly uses NLP technology to extract semantic meaning from the knowledge graph of software service domain.Then,the user feature words in the user portrait are extended by synonym forest and semantic matching is carried out with nodes in the knowledge graph.Finally,the nodes are expanded through the knowledge graph path,and the trans-method is used to supplement the knowledge graph relationship to form individual knowledge graph with user characteristics.（3）Combining the similarity based on project entity,the similarity based on knowledge graph and the similarity based on user score,a hybrid recommendation method of fusing knowledge graph which fully considers user context information is proposed.In terms of accuracy and novelty of recommendation results,this method is superior to the classical collaborative filtering method and the recommendation method that only integrates user images.In this thesis,through the establishment of user portrait,and based on the constructed user portrait,the user node selection,pruning,reconnection and other operations of the domain knowledge graph were conducted to form a personal knowledge graph with user characteristics,so as to recommend software services for users.Through experimental verification,the proposed method alleviates the cold start problem existing in the classical collaborative filtering algorithm,and improves the accuracy and novelty of the recommendation results.