Page 71 - 《软件学报》2025年第5期

胡强 等: 融合潜在联合词与异质关联兼容的 Web API 推荐 1971

由图 9–图 11 可见, 除 K=7 时 Recall 与 Precision 分别在 6 个词与 4 个词中出现轻微波动外, 随潜在特征词融
入数量增加, Recall、Precision 与 NDCG 指标随之提升, 在词数为 6 时到达峰值. 当超过 6 个词, 模型效果会出现
下降趋势. 上述趋势与服务描述文本较短有关, 当融入的场景特征词过多时, 会影响功能特征的语义表达强度. 基
于此, 将潜在特征词提取数量统一设置为 6.

5 结束语

为提升 Mashup 服务的组件 Web API 推荐质量, 提出一种融合潜在联合词与异质关联兼容的 Web API 推荐
方法. 设计了融入场景契合度的服务描述特征词提取模型 SYAKE, 为 Mashup 服务需求和 Web API 提取表示应用
场景的潜在联合词, 借助潜在联合词提升了功能匹配的计算精度. 在此基础上, 借助协同过滤思想, 为 Mashup 服
务需求构建高质量的候选组件 Web API 集合. 构建了一种利用异质关联兼容提升 Web API 推荐质量的方法, 相比
当前流行的以协作兼容提升 Web API 推荐的方法, 异质关联兼容能够有效提升推荐质量和多样性. 实验证明, 本
文方法在推荐的精确度、召回率和多样性等指标上显著优于对比方法.
未来研究工作主要是挖掘更多类型的 Web API 关联, 如 API 调用者关联、地理位置关联等, 以进一步改善关
联兼容对推荐质量的提升效果. 需要注意的是在后续 Web API 的关联挖掘中, 需要分析不同类型的关联对 Web
API 推荐质量评价指标的差异化影响, 以期将有利于提高推荐质量的关联纳入到推荐方法中. 同时, 关联密度对推
荐质量的影响问题也需要加以关注和探讨.

References:
[1] Sang CY, Deng XY, Liao SG. Mashup-oriented Web API recommendation via full-text semantic mining of developer requirements. IEEE
Trans. on Services Computing, 2023, 16(4): 2755–2768. [doi: 10.1109/TSC.2023.3245652]
[2] Lu JW, Zhao W, Zhang YM, Liang QH, Xiao G. TWE-NMF topic model-based approach for Mashup service clustering. Ruan Jian Xue
Bao/Journal of Software, 2023, 34(6): 2727–2748 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/6508.htm [doi: 10.
13328/j.cnki.jos.006508]
[3] Lu JW, Wu H, Zhang YM, Liang QH, Xiao G. Mashup service clustering method via integrating functional semantic association
calculation and density peak detection. Chinese Journal of Computers, 2021, 44(7): 1501–1516 (in Chinese with English abstract). [doi:
10.11897/SP.J.1016.2021.01501]
[4] Qi LY, Song HB, Zhang XY, Srivastava G, Xu XL, Yu S. Compatibility-aware Web API recommendation for Mashup creation via
textual description mining. ACM Trans. on Multimedia Computing, Communications, and Applications, 2021, 17(1s): 20. [doi: 10.1145/
3417293]
[5] Qi LY, He Q, Chen FF, Zhang XY, Dou WC, Ni Q. Data-driven Web APIs recommendation for building Web applications. IEEE Trans.
on Big Data, 2022, 8(3): 685–698. [doi: 10.1109/TBDATA.2020.2975587]
[6] Chen NX, Cardozo N, Clarke S. Goal-driven service composition in mobile and pervasive computing. IEEE Trans. on Services
Computing, 2018, 11(1): 49–62. [doi: 10.1109/TSC.2016.2533348]
[7] Qi LY, He Q, Chen FF, Dou WC, Wan SH, Zhang XY, Xu XL. Finding all you need: Web APIs recommendation in Web of things
through keywords search. IEEE Trans. on Computational Social Systems, 2019, 6(5): 1063–1072. [doi: 10.1109/TCSS.2019.2906925]
[8] Gu Q, Cao J, Peng QY. Service package recommendation for Mashup creation via Mashup textual description mining. In: Proc. of the
2016 IEEE Int’l Conf. on Web Services. San Francisco: IEEE, 2016. 452–459. [doi: 10.1109/ICWS.2016.65]
[9] Li HC, Liu JX, Cao BQ, Tang MD, Liu XQ, Li B. Integrating tag, topic, co-occurrence, and popularity to recommend Web APIs for
Mashup creation. In: Proc. of the 14th IEEE Int’l Conf. on Services Computing. Honolulu: IEEE, 2017. 84–91. [doi: 10.1109/SCC.2017.19]
[10] Zhang N, Wang J, Ma YT. Mining domain knowledge on service goals from textual service descriptions. IEEE Trans. on Services
Computing, 2020, 13(3): 488–502. [doi: 10.1109/TSC.2017.2693147]
[11] Tang MD, Xie FF, Liang W, Xia YM, Li KC. Predicting new composition relations between Web services via link analysis. Int’l Journal
of Computational Science and Engineering, 2019, 20(1): 88–101. [doi: 10.1504/IJCSE.2019.103256]
[12] Su C, Huang DL. Hybrid recommender system based on deep learning model. Int’l Journal of Performability Engineering, 2020, 16(1):
118–129. [doi: 10.23940/ijpe.20.01.p13.118129]
[13] Botangen KA, Yu J, Sheng QZ, Han YB, Yongchareon S. Geographic-aware collaborative filtering for Web service recommendation.