PAREI: A progressive approach for Web API recommendation by combining explicit and implicit information

Mashup is an application with specific functions by combining Web APIs that can provide services or data on the Internet, thus avoiding the behavior of repeatedly building wheels. Recommending suitable Web APIs in the vast number of Web APIs on the Internet for Mashup developers has become a challen...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Information and software technology Ročník 162; s. 107269
Hlavní autoři: Wang, Ye, Zhou, Aohui, Huang, Qiao, Wang, Xiaoyang, Jiang, Bo
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier B.V 01.10.2023
Témata:
Deep learning
Explicit information
Implicit information
Web API recommendation
Deep learning
Explicit information
Web API recommendation
Implicit information
ISSN:0950-5849, 1873-6025
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Mashup is an application with specific functions by combining Web APIs that can provide services or data on the Internet, thus avoiding the behavior of repeatedly building wheels. Recommending suitable Web APIs in the vast number of Web APIs on the Internet for Mashup developers has become a challenging problem. Previous studies often fail to fully exploit and effectively synthesize various types of information between Web APIs and Mashups. This work proposes a Web API recommendation approach - PAREI by combining both explicit and implicit information to progressively optimize the recommendation results. First, PAREI uses the explicit structural information between Mashups and Web APIs to construct the Call Relationship Network (CRN). Second, PAREI calculates explicit semantic similarities between developer’s requirement and Mashups to obtain candidate Mashup nodes in CRN. Then PAREI further mines the implicit structural information between Mashups. A combined similarity score for each Mashup node is calculated. Finally, PAREI uses CRN to obtain candidate Web APIs related to candidate Mashup nodes, and integrates implicit semantic information of Web APIs with combined scores of corresponding Mashups, so as to obtain Top-K Web APIs. Comparison experiments show that PAREI has significantly improved the Recall, Precision, and MAP metrics compared with other approaches. Ablation experiments show that different types of information play various roles in Web API recommendation, and different combination modes have different effects on the recommendation results. This work constructs the PAREI model, which combines explicit and implicit information to obtain Web API recommendation results through a progressive strategy. According to the experiment results, we believe that the PAREI approach can help Mashup developers to find demanded Web APIs rapidly and accurately. •This work proposes PAREI for Web API recommendation based on deep learning.•PAREI mines and combines both explicit information and implicit information.•PAREI is evaluated through a set of experiments with the real dataset from PWeb.•The role of each type of information in Web API recommendation is revealed.
ISSN:0950-5849
1873-6025
DOI:10.1016/j.infsof.2023.107269