Web APIs provide enterprises with a new way of driving innovations of new technology with limited resources. API recommendations greatly alleviate the selection burdens of enterprises in identifying potential useful APIs to meet their business demands. However, these approaches disregard the privacy leakage risk in cross-platform collaboration and the popularity bias in recommendation. To address these issues, first, we introduce MinHash, an instance of locality-sensitive hashing, into a collaborative filtering technique and propose a novel, privacy-enhanced, API recommendation approach. Second, we present a simulation algorithm to analyze the popularity bias in API recommendation. Third, we mitigate popularity bias by improving the novelty of recommendation results with an adaptive reweighting mechanism. Last, comprehensive experiments are conducted on a real-world dataset collected from ProgrammableWeb. Experimental results show that our proposed approach can effectively preserve usage data privacy and mitigate popularity bias at a minimum cost in accuracy.

Web API为企业提供了一种利用有限资源推动新技术创新的新方式。API推荐极大地减轻了企业在识别潜在有用的API以满足其业务需求时的选择负担。然而，这些方法忽视了跨平台协作中的隐私泄露风险和推荐中的流行度偏差。为了解决这些问题，首先，我们将 MinHash（局部敏感哈希的一个实例）引入到协同过滤技术中，并提出了一种新颖的、隐私增强的 API 推荐方法。其次，我们提出了一种模拟算法来分析 API 推荐中的流行度偏差。第三，我们通过自适应重新加权机制提高推荐结果的新颖性，从而减轻流行度偏差。最后的，在从 ProgrammableWeb 收集的真实数据集上进行了全面的实验。实验结果表明，我们提出的方法可以有效地保护使用数据隐私并以最小的准确性成本减轻流行度偏差。