Web API Recommendation Approaches Based On Topic Model And Factorization Machine

The purpose of this paper is to reomend a set of Web API to help solving the problem of Mashup creation,based on the nature language description of user's requirements.However,a series of question make the realization of this goal become more and more difficult,such as the rapid increase about the number and varity of Web APIs,the description documents of Web services are unstructure,and there are many Web APIs that have similar functions but with different performance characteristics.This paper considers that the technology of topic model can improve understanding for functional information which contained in the Mashup requirement text,and obtain its latent topic distribution vector,and mine the potential semantic relationship between the Mashup requirement text and the Web API description document.Furthermore,Factorization machine model can input various supplementary information,which can effectively reduce the sparseness of the Web API historical call matrix in the traditional collaborative filtering and matrix decomposition methods,and optimize the way of feature combination.This paper,based on the above analysis,presents a novel recommendation appoach,which fusion topic model and factorization machine model to predict or recommend Top-N Web APIs auxiliary the target Mashup creation.The main research methods of this paper are as bellows:(1)A Web API recomendation that integrate tag,topic,co-occurrence and popularity(TR-FM).This method firstly computes the tag-imilarity of Web services(Mashups or Web APIs)by two step,expanding the tags of Web Services and calculating the important weight of the expanded tags for the corresponding Web Services.Secondly,TR-FM calculate the text-similarity between Web services via the latent topic distribution vector derive by RTM model.Then,TR-FM calculates the popularity of Web API by combining the history iovacation times and category information,and measures the co-occurrence of Web API via using classical Jaccard similarity coefficient.Finally,the Top-N Web API collection is recommended by using the factorization machine model to integrate the above features.The experimental results show that TR-FM has good performance in terms of precision,recall and F-measure(2)A novel Web API Recommendation method based on HDP topic Model and Factorization machine(HDP-FM).This method exploits HDP topic model to extract the optimal topic distribution vector of Web Services(Mashup or Web API).Then,HDP-FM utilizes enhanced cosine similarity formula to compute the similarity between Web services.Finally,HDP-FM takes the advantage of the factorization machines model to score and sort Web APIs which makes the similarity between Mashups,the similarity between Web APIs,the popularity of Web APIs and the co-occurrence of Web APIs as inputs.To verify the performance of the HDP-FM method,we conducted a series of experiments on a real dataset crawled from the ProgrammableWeb platform.The results show that the HDP-FM method has a good performance over others in term of precision,recall,F-measure and NDCG@N.