陈佳佩, 武浩, 秦绍伟, 彭伟乐, 徐立. 面向Mashup的质量感知Web API推荐[J]. 云南大学学报(自然科学版), 2022, 44(4): 688-697. doi: 10.7540/j.ynu.20210440
引用本文: 陈佳佩, 武浩, 秦绍伟, 彭伟乐, 徐立. 面向Mashup的质量感知Web API推荐[J]. 云南大学学报(自然科学版), 2022, 44(4): 688-697. doi: 10.7540/j.ynu.20210440
CHEN Jia-pei, WU Hao, QIN Shao-wei, PENG Wei-le, XU Li. Quality-aware Web API recommendation for Mashup[J]. Journal of Yunnan University: Natural Sciences Edition, 2022, 44(4): 688-697. DOI: 10.7540/j.ynu.20210440
Citation: CHEN Jia-pei, WU Hao, QIN Shao-wei, PENG Wei-le, XU Li. Quality-aware Web API recommendation for Mashup[J]. Journal of Yunnan University: Natural Sciences Edition, 2022, 44(4): 688-697. DOI: 10.7540/j.ynu.20210440

面向Mashup的质量感知Web API推荐

Quality-aware Web API recommendation for Mashup

  • 摘要: 近年来,网络中Web API的数量日益增多,如何面向Mashup应用推荐合适的高质量Web API已成为研究的热点问题,现有的方法忽略了Web API质量信息对推荐的影响从而制约了其性能. 深度学习技术为进一步提高Web API推荐的准确性提供了新的解决方案,如何利用Web API质量信息并结合深度网络模型进行高精度推荐也成为关键问题. 为此,提出了一种Web API质量感知的深度推荐模型. 首先,使用BERT预训练模型作为文本编码器对Mashup和Web API的文本描述特征进行提取;然后,借助自注意力机制对Web API的质量信息进行融合,并利用所得的Web API质量增强特征进行推荐. 基于真实数据集的实验结果表明,对比基线方法,该模型在Web API 推荐任务的top-1准确率、召回率和归一化折损累积增益指标上分别提高了3.97%、3.45%和3.97%.

     

    Abstract: The number of Web APIs on the Internet has increased in recent years. How to recommend a suitable high-quality Web API for Mashup applications has become a hot research topic. Ignoring the influence of Web API quality information on recommendations has restricted the performance of the existing technical methods. Deep learning technology provides a new solution to further improve the accuracy of Web API recommendations. How to use Web API quality information combined with deep network models for high-precision recommendations has also become a key issue. To this end, this paper proposes a Web API quality-aware deep recommendation model, which combines deep neural networks and quality information modelling to improve the accuracy of Web API recommendations. First, we use the BERT pre-training model as a text encoder to extract the text description features of Mashup and Web API, then utilize the self-attention mechanism to fuse Web API quality information, and finally make recommendations based on the obtained Web API quality enhancement features. Compared with the baseline model of Web API tasks based on real dataset, the experimental results prove that our method improves the top-1 accuracy, recall, and NDCG indicators by 3.97%, 3.45% and 3.97% respectively

     

/

返回文章
返回