API凭借其松耦合及跨平台跨云的特性，正在成为一种重要的互联网技术和思维模式。随着移动互联网，物联网和云计算技术的迅猛发展，将涌现大量的公用及私有的API, 不仅为用户带来更多的服务选择和更好的用户体验，同时也为面向API的发现和聚合带来了挑战。目前，对于API的研究国内外都刚刚起步，特别是对API数量巨大，网络环境复杂，冷启动及数据稀疏等问题的研究尚未有可靠的解决方案。本项目的研究内容如下：1）在现有的API发现研究基础上，引入社交媒体数据，加强基于语义和功能性的API发现准确率；2）引入API相关的多源数据集，基于从中提取的异构特征，利用回归模型预测API的服务质量（如热门程度等），从而解决大量API涌现所带来的冷启动、数据稀疏等问题；3）研究在移动互联网和海量API环境下的API聚合问题，通过引入流形学习等方法，预测、优化、推荐更好的聚合API给用户和开发者。

Due to the character of loose coupling, cross-platform, and cross-cloud, API is becoming an important Internet technique and thinking model. With the rapid development of mobile Internet, IoT, and cloud computing, there will be a mass of open/private APIs, which not only provide more service choices and better user experience, but also pose challenges to the API oriented discovery and mashup creation. Currently, there are few papers about API research, especially for the scenario in which there are massive APIs, complicated network environment, cold-start, and data sparsity problems. This proposal tries to tackle above problems via the following research: 1) Based on current research of API discovery, we propose to import the API related social media data, for the purpose of improving the accuracy of semantic and functional based API discovery; 2) We propose to import more API related data sources, extract more heterogeneous features, and then predict the service quality of API to tackle the problems caused by cold-start and data sparsity; 3) We propose to utilize manifold learning technique to predict, optimize, and recommend better API mashup to users and developers.