| 摘要: |
| 服务描述中包含的应用场景信息有限, 使得以功能相似度计算为主的Mashup服务组件Web API推荐与需求预期常存在差异, 功能匹配精确度有待进一步提高. 部分研究者虽利用Web API的协作关联提升推荐兼容性, 但忽视了功能关联对Mashup服务创建的负反馈影响, 从而限制了推荐多样性的提升. 为此, 提出一种融合潜在联合词与异质关联兼容的Mashup服务的组件Web API推荐方法. 该方法为Mashup需求和Web API提取潜在应用场景联合词并融入到功能向量的生成中, 进而提高二者功能相似度的匹配精确度, 以获得高质量的候选组件Web API集合. 将功能关联与协作关联建模为异质服务关联, 并利用异质关联兼容替代传统方法中的协作兼容, 以提升Web API的推荐多样性. 相较于对比方法, 所提方法在评价指标Recall、Precision和NDCG上分别提升了4.17%–16.05%, 4.46%–16.62%与5.57%–17.26%, 多样性指标ILS降低了8.22%–15.23%. 冷启动Web API推荐的Recall与Precision指标值分别为非冷启动Web API推荐的47.71%和46.58%. 实验结果表明所提方法不仅提升了Web API推荐质量, 而且对冷启动Web API具有很好的推荐效果. |
| 关键词: Mashup服务 异质关联 Web API推荐 多样性 |
| DOI：10.13328/j.cnki.jos.007185 |
| 分类号:TP311 |
| 基金项目:国家自然科学基金(61973180); 国家重点研发计划(2023YFF0612100); 山东省自然科学基金(ZR2021MF092); 山东省重点研发计划(软科学)(2023RKY01009); 云南省服务计算重点实验室开放课题(YNSC23116) |
|
| Web API Recommendation by Fusing Latent Related Words and Heterogeneous Association Compatibility |
|
HU Qiang1,2, QI Hao-Quan1, LI Hao-Jie1, DU Jun-Wei1
|
|
1.College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, China;2.Yunnan Key Laboratory of Service Computing, Kunming 650221, China
|
| Abstract: |
| The service descriptions provide limited information about application scenarios, creating a gap between Mashup service component Web API recommendations based on functional similarity calculation and desired expectations. Consequently, there is a need to enhance the accuracy of function matching. While some researchers utilize collaborative associations among Web APIs to enhance recommendation compatibility, they overlook the adverse effects of functional associations on Mashup service creation, thereby limiting the enhancement of recommendation diversity. To address this issue, this study proposes a Web API recommendation method for Mashup service components that integrates latent related words and heterogeneous association compatibility. The study extracts latent related words associated with application scenarios for both Mashup requirements and Web APIs, integrating them into the generation of function vectors. By enhancing the accuracy of functional similarity matching, it obtains a high-quality candidate set of Web API components. Function association and collaboration association are modeled as heterogeneous service association. The study utilizes heterogeneous association compatibility to replace collaboration compatibility in traditional methods, thus enhancing the recommendation diversity of Web APIs. In comparison, the proposed approach demonstrates improvements in evaluation indicators, with Recall, Precision, and NDCG enhanced by 4.17% to 16.05%, 4.46% to 16.62%, and 5.57% to 17.26%, respectively. Additionally, the diversity index ILS is reduced by 8.22% to 15.23%. The Recall and Precision values for cold-start Web API recommendation are 47.71% and 46.58% of those for non-cold-start Web API recommendation, respectively. Experimental results demonstrate that the proposed method not only enhances the quality of Web API recommendation but also yields favorable results for cold-start Web API recommendations. |
| Key words: Mashup service heterogeneous association Web API recommendation diversity |
