.png){kind=icon}
.png){kind=icon}
Return
CoWAR: A General Complementary Web API Recommendation Framework Based on Learning Model
.png){kind=icon}
PRE
AI
DOI：10.1109/TR.2026.3667714.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
With the rapid advancement of service computing technologies, the proliferation of Web application programming interfaces (APIs) on the Internet has increased exponentially. However, selecting the most suitable APIs for Mashup creation from this extensive pool presents a significant challenge for users. Numerous Web API recommendation methods have been developed to address this issue, aiming to simplify the complex selection process. Despite these advancements, there has been limited research on the recommendation of complementary functions. In response, we propose complementary Web API recommendation (CoWAR), a comprehensive framework for recommending complementary Web APIs tailored to Mashup creation, based on the Web APIs previously selected by users. Specifically, we introduce a data labeling algorithm that generates a labeled dataset using Mashup–API interactions derived from historical Mashups and Web APIs. Furthermore, we utilize the Sentence-bidirectional-encoder-representations-from-transformers model to generate representation vectors of Web APIs from their functional descriptions. Subsequently, a self-attentional neural factorization machines model is employed to train the CoWAR model on the labeled dataset, utilizing the Web APIs’ representation vectors. An attention mechanism is integrated into CoWAR to identify varying complementary weights between the selected Web APIs and candidate Web APIs, thereby enhancing recommendation performance. To the best of our knowledge, this is the first work to address the complementary function recommendation problem using a learning-based approach. Experimental validation on a real-world dataset demonstrates the effectiveness of the proposed framework, showing that the learning model outperforms both traditional machine learning-based models and several deep learning-based models.
Keywords:
Complementary function
learning model
mashup
service recommendation
web application programming interface (API)
Journal
IF:
5.7
Papers:
2.7K
Citations:
8.5K