Mining Stable Communities in Temporal Networks by Density-Based Clustering

Qin, Hongchao; Li, Rong Hua; Wang, Guoren; Huang, Xin; Yuan, Ye; Yu, Jeffrey Xu

doi:10.1109/tbdata.2020.2974849

Cited by 14 publications

(13 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Appel et al (2019) used a shared factorization model that can account for links, weights, and temporal analysis in graphs to detect communities. Qin et al (2020) proposed a density-based clustering algorithm to detect stable communities over time in a dynamic graph. In this paper, we used the algorithm proposed by Boudebza et al (2020), a link stream-based approach, to detect stable communities at multiple temporal scales without redundancy instead of choosing an arbitrarily temporal scale and to perform temporal community detection on all time steps in parallel.…”

Section: Dynamic Community Detectionmentioning

confidence: 99%

Homophilic and Heterophilic Characteristics Shaping Community Formation in Human Mobility Networks During Extreme Weather Response

Lee¹,

Namburi²,

Xiao³

et al. 2022

Preprint

View full text Add to dashboard Cite

Community formation in socio-spatial human networks is one of the important mechanisms for ameliorating hazard impacts of extreme weather events. Research is scarce regarding latent network characteristics shaping community formation in human mobility networks during natural disasters. We examined human mobility networks in Harris County, Texas, in the context of the managed power outage forced by Winter Storm Uri to detect communities and to evaluate latent characteristics in those communities. We examined three characteristics in the communities formed within human mobility networks: hazard-exposure heterophily, socio-demographic homophily, and social connectedness strength. The analysis results show that population movements were shaped by socio-demographic homophily, heterophilic hazard exposure, and social connectedness strength. The results also indicate that a community encompassing more high-impact areas would motivate population movements to areas with weaker social connectedness. Hence, the findings reveal important characteristics shaping community formation in human mobility networks in hazard response. Specific to managed power outages, formed communities are spatially co-located, underscoring a best management practice to avoid prolonged power outages among areas within communities, thus improving hazardexposure heterophily. The findings have implications for power utility operators to account for the characteristics of socio-spatial human networks when determining the patterns of managed power outages.

show abstract

Section: Dynamic Community Detectionmentioning

confidence: 99%

Homophilic and Heterophilic Characteristics Shaping Community Formation in Human Mobility Networks During Extreme Weather Response

Lee¹,

Namburi²,

Xiao³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…To show the effectiveness of finding communities in dynamic or temporal networks, we compare our (θ, k)-core reliable community (CRC) with the Persistent Community (PC) and Stable Community (SC) proposed by Li [9] and Qin [10], respectively. To do this, we select five graph instances of Reddit dataset and return the largest community C obtained by SC, PC, and our CRC.…”

Section: F Effectiveness Evaluation Of Query Processingmentioning

confidence: 99%

“…Although there are two similar existing works [9], [10] to identify meaningful communities over time, this work is still desirable due to the different research challenges. In [9], Li et al defined the persistent community search as the maximal k-core where each vertex's accumulated degree meets the k-core requirement within a time interval.…”

Section: Introductionmentioning

confidence: 99%

“…They designed a novel temporal graph reduction algorithm and searched the maximum persistent community utilizing pruning and bounding techniques, which takes exponential complexity. In [10], Qin et al proposed the stable communities by first selecting the centroid vertices where each centroid vertex has a certain number of neighbors with the desired similarity, and the star-shape of the centroid vertex and its neighbors exists frequently in a period of time; and then clustering the network vertices into stable groups based on the selected centroids. However, the persistent community in [9] did not consider the weight of the edge and its time complexity is too high for dealing with large-scale networks.…”

Section: Introductionmentioning

confidence: 99%

“…However, the persistent community in [9] did not consider the weight of the edge and its time complexity is too high for dealing with large-scale networks. The stable community in [10] ignored the continuity of the cohesive structure and failed to maximize member engagement.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Reliable Community Search in Dynamic Networks

Tang¹,

Li²,

Haldar³

et al. 2022

Preprint

View full text Add to dashboard Cite

Local community search is an important research topic to support complex network data analysis in various scenarios like social networks, collaboration networks, and cellular networks. The evolution of networks over time has motivated several recent studies to identify local communities from dynamic networks. However, they only utilized the aggregation of disjoint structural information to measure the quality of communities, which ignores the reliability of communities in a continuous time interval. To fill this research gap, we propose a novel (θ, k)core reliable community (CRC) model in the weighted dynamic networks, and define the problem of the most reliable community search that couples the desirable properties of connection strength, cohesive structure continuity, and the maximal member engagement. To solve this problem, we first develop an online CRC search algorithm by proposing a definition of eligible edge set and deriving the eligible edge set based pruning rules. After that, we devise a Weighted Core Forest-Index and index-based dynamic programming CRC search algorithm, which can prune a large number of insignificant intermediate results according to the maintained weight and structure information in the index, as well as the proposed upper bound properties. Finally, we conduct extensive experiments to verify the efficiency of our proposed algorithms and the effectiveness of our proposed community model on eight real datasets under different parameter settings.

show abstract