| Software reuse is that developers usually make full use of the knowledge and experience accumulated in the past development.Software reuse went through three stages,they are partial reuse based on code blocks,reuse of small scale based on component technology and reuse of large scale based on open source software.At the same time,more and more people take part in the open source movement at present,which enrich the reusable resources in the open source world and provide a wider choice for software reuse.However,the characteristics that the number of open source software is very large and grows fast make its scale and diversity far beyong the developers' cognitive range.Usually,developers face enormous challenges when they pinpoint and reuse open source software they need quickly and accurately even nicely.According to these challenges mentioned above,this paper regard that open source recommendation is a new mechanism of software reuse application under open souce mode as a new idea.Based on this idea,two kinds of recommendation algorithm are studied and a prototype recommendation system is implemented in this paper:1?Similar software recommendation algorithm research based on reuse network.In GitHub open source hosting community,software projects are often highly open and free management.They have no label information and may not have description information.So it is very difficult to identify the differences between similar software through text attributes.Therefore,this paper builds reuse network and analyzes this network from reuse relation perspective-a new perspective and proposes similar software recommendation algorithm based on reuse network.In this algorithm,reuse relation and TF-IDF value are regarded as characteristic and weight respectively.It calculates the similarity between software projects with cosine rule.At the same time,this paper makes the real software projects in GitHub as a data set to do experimental evaluation.Evaluation results show that the recommendation algorithm has a good effect,and they can be used as a supplement for the results of searching.2?Combined software recommendation algorithm research based on co-occurrence network.In the process of software development based on software reuse,besides helping developers pinpoint and reuse open source software project they need quickly and accurately even nicely,we also need to let developers understand what software projects can be used together and try to avoid software projects which used together will have conflicts.But the scale of the reuse relation of open source software is very large and relatively sparse,which makes the traditional concept of frequent item sets and association rules under the background of software reuse in not ideal.Therefore,this paper builds cooccurrence network,analyzes this network from co-occurrence relation perspective and proposes combined software recommendation based on co-occurrence network.In this algorithm,based on co-occurrence network,the attributes of nodes and edges and circle finding are regarded as computing object and optimizing method respectively.It discovers the co-occurrence relation between software projects effectively.And the heuristic thought is the main strategy to achieve the recommendation of combined software.Similarly,this paper makes the same date set to do experimental evalution and also achieves good recommendation effect.3?design and implement open source software integrated recommendation system.in order to transform the recommendation algorithm into service,this paper designs and implements SC-OSSREC open source software integrated recommendation system subsequently.This recommendation system has four modules,including data collecing,data processing,algorithm implementing,results displaying.At present,this recommendation system can recommend corresponding software project lists for more than 120,000 open source software projects,and enhance the effect of the recommendation according to the user feedback by doing dynamic adjustment. |
PDF Full Text Request