HeteRecom

Shi, Chuan; Zhou, Chong; Kong, Xiangnan; Yu, Philip S.; Liu, Gang; Wei, Bai

doi:10.1145/2339530.2339778

Cited by 52 publications

(6 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Meta-path selection is necessary because trivial meta-paths may include noise and cause extra computational cost. Selecting meta-paths from the HIN can be carried out in three ways: (1) by using users' domain knowledge, which is straightforward but relies heavily on experts' experience and effort [31]; (2) by incorporating all meta-paths within a given length, which avoids information loss but brings noise and leads to low efficiency [32]; and (3) by selecting metapaths with automatic mechanisms. For example, Meng et al propose a greedy algorithm to select meta-paths based on user-provided node pairs [33].…”

Section: Hin Mining For Recommendationsmentioning

confidence: 99%

Incorporating heterogeneous information in deep learning with informative meta-paths for community recommendations

Deng

Han

2021

Journal of Information Science

View full text Add to dashboard Cite

Communities of interest promote knowledge sharing and discovery in social network platforms. However, platform users face difficulties of finding suitable communities, given their increasing number. Although recommendations have been proposed to help users find communities of interest, these methods ignore or exclude heterogeneous interactions between users and communities. In addition, widely used meta-paths help capture the complex semantic relation among entities but heavily rely on domain knowledge. In this study, we propose a novel recommendation model based on informative meta-path discovery in heterogeneous information networks and deep learning. Users, communities, relevant items and their relations are considered as entities in a heterogeneous information network, from where informative meta-paths are extracted on the basis of information theory to measure user-community similarities. Finally, similarities are incorporated in a deep learning model to predict whether target users join candidate communities. The proposed recommendation model is evaluated and compared against baseline methods using two data sets. Results demonstrate the superior performance of the present model in terms of precision, recall and F score.

show abstract

Section: Hin Mining For Recommendationsmentioning

confidence: 99%

Incorporating heterogeneous information in deep learning with informative meta-paths for community recommendations

Deng

Han

2021

Journal of Information Science

View full text Add to dashboard Cite

show abstract

“…In the following, two different similarity measures used in this work are described, Path-Sim and HeteSim. Relevance recommendation aims at identifying multi-typed items related to a query object in terms of meaningful semantic relationships [3]. Finally ranking in HINs allows to identify the most relevant multi-typed nodes according to the selected meta-path [2].…”

Section: Background On Hins and Meta-pathsmentioning

confidence: 99%

“…The proposed algorithm is related to the relevance recommendation algorithm HeteRecom [3], which automatically identifies relevant semantic meta-paths in HINs, starting from a query node. However, most of the possible semantic metapaths are not meaningful in this scenario, due to the high uncertainty in the EG links.…”

Section: Evidence Recommendation Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Evidence recommendation in forensics based on cyclic meta-paths in heterogeneous information networks

Koblents

Álvarez

2017

8th International Conference on Imaging for Crime Detection and Prevention (ICDP 2017)

View full text Add to dashboard Cite

In this article we propose a novel approach for evidence recommendation in digital forensic applications. The so-called Evidence Graph (EG) is a Heterogeneous Information Network (HIN) constructed from entities and relationships extracted from large collections of low quality video footage, collected by surveillance cameras in relation to a crime. The EG has been developed in the context of the EU-funded project LASIE. It presents multiple types of nodes and edges and a generic network schema in which all types of nodes can be connected. In this work, relevant cyclic meta-paths are identified and an iterative scheme for evidence recommendation is proposed and evaluated on real data.

show abstract

“…Zhang et al have formalized the recommendations over a heterogeneous social network as a ranking problem and proposed a random walk model to estimate the importance of each object in the heterogeneous network. As an alternative to conventional recommendation systems that recommend similar objects of the same type, Shi et al have developed HeteRecom to evaluate the relatedness between any‐typed objects using a path‐based relevance measure, which can recommend similar objects of the same type and also related objects of different types.…”

Section: Experimental Developments and Applicationsmentioning

confidence: 99%

Clustering on heterogeneous networks

Huang

Gao

2014

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Objects that are interrelated with each other are often represented as homogeneous networks, in which objects are of the same entity type and relationships between objects are of the same relationship type. However, heterogeneous information networks, composed of multiple types of objects and/or relationships, are ubiquitous in real life. Mining heterogeneous information networks is a new and promising field of research in data mining, and clustering is an important way to identify underlying patterns in data. Although clustering on homogeneous networks has been studied for several decades, clustering on heterogeneous networks has been explored only recently. However, some progress has already been made with respect to this theme, ranging from algorithms to various related applications. This paper presents a brief summary of current research regarding heterogeneous network clustering and addresses some promising research directions. First, it presents a formalized definition and two important aspects of heterogeneous information networks to elaborate why clustering on heterogeneous networks is of significance. Then, this review provides a concise classification of existing heterogeneous network clustering algorithms based on their methodological principles. In addition, it discusses experimental developments and applications of heterogeneous network clustering. The paper addresses several open problems and critical issues for future research. WIREs Data Mining Knowl Discov 2014, 4:213–233. doi: 10.1002/widm.1126 This article is categorized under: Algorithmic Development > Structure Discovery Technologies > Computational Intelligence Technologies > Structure Discovery and Clustering

show abstract

HeteRecom

Cited by 52 publications

References 8 publications

Incorporating heterogeneous information in deep learning with informative meta-paths for community recommendations

Incorporating heterogeneous information in deep learning with informative meta-paths for community recommendations

Evidence recommendation in forensics based on cyclic meta-paths in heterogeneous information networks

Clustering on heterogeneous networks

Contact Info

Product

Resources

About