The real-world systems usually contain different types of components that interact with each other. Most existing work models these interaction systems as homogeneous information networks, which does not considerthe heterogeneous interaction relationships among objects, resulting in lots of information loss. In recent years, more researchers model these interaction data as heterogeneous information networks (HINs) and conduct knowledge discovery based on the comprehensive structural information and rich semantic information in HINs. Specifically, with the advent of the era of big data, HINs naturally merge heterogeneous data sources, which make it an important way to solve the variety of big data. Therefore, heterogeneous information network analysis has quickly become a hot spot in data mining research and industrial applications. This article provides a comprehensive overview of heterogeneous information network analysis and applications. In addition to the basic concepts in heterogeneous information networks, the focus of this article is on the latest research progress in meta-path based data mining, heterogeneous information networks representation learning, and practical applications of heterogeneous information networks. In the end, this article points out the possible directions of future development.