大数据的产生一方面为推荐系统提供了丰富的学习样本，另一方面也呈现出新的结构特征，这些特征对推荐系统提出了更高的要求。本项目研究的目的是:创建新的多源异构数据张量补全的隐马尔科夫模型推荐方法，解决大数据背景下协同过滤推荐系统的冷启动问题、数据稀疏问题、实时性问题。拟从显性与隐性双数据视角分析问题，以揭示多源异构数据内在的融合机制为核心，分析多源异构数据的内在联系以及可能的映射关系。通过建立用户隐性轨迹数据的隐马尔科夫动态推荐方法捕获用户隐性偏好的时间上下文信息，并结合基于显性数据的张量补全模型优化推荐结果。本项目研究成果不仅对完善数据补全与融合理论、协同过滤推荐方法具有重要的价值，并且为交通智能诱导中实时路径推荐应用提供技术支撑。

The generation of big data provides rich learning samples for the recommendation system. It also presents new structural features, which put forward higher requirements for the recommendation system. The purpose of this project is to develop a new Hidden Markov models (HMM) recommendation method for multi-source heterogeneous data tensor completion, to solve the cold start, data sparsity and real-time problems of collaborative filtering recommendation system under the background of big data. This project intends to analyze the problem from the perspective of explicit and implicit data, focusing on revealing the internal fusion mechanism of multi-source heterogeneous data, and analyzing the internal relationship and possible mapping relationship of multi-source heterogeneous data. The HMM dynamic recommendation method is established to capture the time context information of users' implicit preferences, and the recommendation results are optimized with tensor completion model based on explicit data. The research results of this project not only have important value to improve the theory of data completion and fusion, collaborative filtering recommendation methods, but also provide technical support for the application of path recommendation in traffic intelligent guidance.