Research On Software Recommendation Method Based On Open Source Community And User Behavior

With the software development technology continuously updating iteration,how to improve the efficiency and quality of software development is the first problem in software engineering,software reuse is an important way to solve this problem.In recent years,with the emergence of various open source communities,more and more developers and people in related fields are participating in sharing and learning resources about open source software.For a long time,the open source community has accumulated a large amount of open source software resources.These resources are huge in size,wide in scope and uneven in quality,which make it more difficult for users to find reusable project resources and reduce the accuracy of the recommended results.In an open source community like Gitee,users typically search for items of interest or relevance to their work by entering keywords related to the project's functionality.However,keywords do not fully describe the functional characteristics of a software project.It is difficult for users to select the appropriate keywords for search,so the results obtained by keyword search are inefficient and can not quickly recommend the projects they really need.In summary,users are faced with difficulties and inefficiencies in searching for reusable project resources in the source community.To solve these problems,the following work has been done:(1)Feature extraction based on the description document and source code.Description documents and source code in project warehouse contain information that can be used as feature of project function.Starting from this feature,the text of code and comment information in description documents and source code are discussed in detail,different preprocessing and filtering rules are designed,from which a vocabulary list that can be used as feature of project function is extracted and a project similarity matrix is constructed.(2)Use user behavior as a reference factor for the recommended methods in this paper.User behavior represents the user's personalized needs.This paper builds a user-project matrix based on user behavior,calculates its product of similarity with the project matrix based on descriptive documents and source code,and then sorts to generate software recommendation results.In addition,this method uses the user's positive and negative feedback as a reference factor for the recommended algorithm and then optimizes the results twice.(3)Reasonable design of the system architecture based on the user's needs.The method proposed in this paper is based on Gin,GRPC,Tensort Flow framework,and a set of separate front-end and back-end software recommendation system is built.