A Survey of Heterogeneous Information Network Analysis This paper provides a survey of heterogeneous information network (HIN) analysis. HINs are networks with multiple types of objects and links, which contain richer structure and semantic information compared to homogeneous networks. HINs have become a hot topic in data mining, database, and information retrieval, with many novel data mining tasks explored in such networks, such as similarity search, clustering, and classification. The paper introduces basic concepts of HIN analysis, examines its developments on different data mining tasks, discusses some advanced topics, and points out some future research directions. HINs are defined as directed graphs with object type mapping and link type mapping functions. The network schema describes the meta structure of a network, and meta paths are paths defined on a schema that define composite relations between objects. Meta paths are important for HIN analysis, as they allow for different connection relations with diverse path semantics, which may have an effect on many data mining tasks. HINs are compared with other related concepts, such as homogeneous networks, multi-relational networks, and composite networks. HINs are different from homogeneous networks in that they include different types of nodes or links. They are also different from multi-relational networks, which have only one type of object but more than one kind of relationship between objects. HINs are also different from composite networks, which have users with various relationships and share some common latent interests across networks. The paper also discusses example datasets of HINs, such as multi-relational networks with single-typed object, bipartite networks, star-schema networks, and multiple-hub networks. These networks are used to model real systems with multi-typed interacting objects. The paper also discusses why HIN analysis is important. HINs can fuse more information and contain rich semantics in nodes and links, making them effective for data mining. They can also be used to handle complex big data, as they are a semi-structured representation. HINs can also be used to fuse information across multiple social network platforms, as they can model different types of objects and relations among them. The paper also discusses research developments in HIN analysis, including similarity measure, clustering, classification, link prediction, and ranking. These tasks have been extensively studied in HINs, with many novel methods proposed to handle the challenges of HIN analysis. The paper concludes that HIN analysis is an important and meaningful task, with many future research directions to explore.A Survey of Heterogeneous Information Network Analysis This paper provides a survey of heterogeneous information network (HIN) analysis. HINs are networks with multiple types of objects and links, which contain richer structure and semantic information compared to homogeneous networks. HINs have become a hot topic in data mining, database, and information retrieval, with many novel data mining tasks explored in such networks, such as similarity search, clustering, and classification. The paper introduces basic concepts of HIN analysis, examines its developments on different data mining tasks, discusses some advanced topics, and points out some future research directions. HINs are defined as directed graphs with object type mapping and link type mapping functions. The network schema describes the meta structure of a network, and meta paths are paths defined on a schema that define composite relations between objects. Meta paths are important for HIN analysis, as they allow for different connection relations with diverse path semantics, which may have an effect on many data mining tasks. HINs are compared with other related concepts, such as homogeneous networks, multi-relational networks, and composite networks. HINs are different from homogeneous networks in that they include different types of nodes or links. They are also different from multi-relational networks, which have only one type of object but more than one kind of relationship between objects. HINs are also different from composite networks, which have users with various relationships and share some common latent interests across networks. The paper also discusses example datasets of HINs, such as multi-relational networks with single-typed object, bipartite networks, star-schema networks, and multiple-hub networks. These networks are used to model real systems with multi-typed interacting objects. The paper also discusses why HIN analysis is important. HINs can fuse more information and contain rich semantics in nodes and links, making them effective for data mining. They can also be used to handle complex big data, as they are a semi-structured representation. HINs can also be used to fuse information across multiple social network platforms, as they can model different types of objects and relations among them. The paper also discusses research developments in HIN analysis, including similarity measure, clustering, classification, link prediction, and ranking. These tasks have been extensively studied in HINs, with many novel methods proposed to handle the challenges of HIN analysis. The paper concludes that HIN analysis is an important and meaningful task, with many future research directions to explore.