| The rapid development of Internet technology makes Mashup and other software’s development more and more intelligent.Normally,Mashup developers can create their interested Mashups by economically composing the existing web APIs(Web Application Programming Interface).And for the development convenience,the number of functional web APIs in various service providing platforms(i.e.,Programmable Web.com)are increasingly growing.Nowadays,the big volume of available web APIs as well as their function differences place a heavy burden on selecting a set of appropriate web APIs which developers are really interested in.In this situation,the recommender system has played an important role and various Web API recommendation approaches have also been continuously proposed.However,the existing approaches are still facing some challenging problems.For example,the existing recommendation approaches often ignore the diversity of APIs,which may lead to the redundancy of returning results and the limitation of developers’ options.And the existing recommendation approaches are often with the poor efficiency,which may lead to the time-consuming process and slow recommendation.In view of these issues and challenges,this paper proposes a diversity-driven and compatible Web API recommendation approach named PD-WACR,as well as an efficiency-driven and compatible Web API recommendation approach named PWAR.The specific research contents are summarized as follows.(1)A Web API recommendation approach named PD-WACR is proposed in this paper.In PD-WACR,an improved weighted Web API correlation graph is constructed and Group Steiner Tree search algorithm as well as Simhash algorithm is introduced.It is a recommendation approach which takes a comprehensive consideration of compatibility,popularity and diversity.Specifically,considering the compatibility between APIs and the popularity of APIs simultaneously,an improved and weighted Web API correlation graph is constructed by utilizing the historical Mashup-API composition records.And a new comprehensive indicator of popularity and compatibility is defined.Then based on the functional requirements of developers,the Group Steiner Tree search algorithm is utilized to dynamically search for multiple sets of candidate Web API compositions with the best compatibility and popularity.After that,Simhash and Hamming Distance are introduced to calculate the degree of diversity between Web API compositions,and the Top-k Web API compositions with the best diversity are selected as the final recommendation results.Finally,relevant experiments are carried out in the real dataset PW.The experiment results proved PD-WACR has wide applicability and can effectively improve the diversity of recommendation results.(2)A Web API recommendation approach named PWAR is proposed in this paper.It combines the Group Steiner Tree search algorithm with the Minimum Spanning Tree algorithm and introduces an efficient progressive recommendation mechanism,which can improve the inefficiency of dynamic search when the number of required keywords is large.Specifically,a weighted Web API correlation graph is constructed by utilizing the compatibility information and function information of APIs.Then combining with the correlation graph and the functional requirements of developers,PWAR utilizes the Group Steiner Tree search algorithm and the Minimum Spanning Tree algorithm to quickly generate a feasible solution which is compatible and can meet the requirements.For each intermediate solution that is generated in the search process,PWAR expands it to a new optimized solution with better compatibility than the previous returning one until the optimal solution is found.Finally,experiment results show that PWAR can greatly improve the efficiency of recommendation with the guarantee of high accuracy and the performance is significantly better than other approaches. |
