Network Clustering Method Research Articles

Clique-Based Method for Social Network Clustering

In this article, we develop a clique-based method for social network clustering. We introduce a new index to evaluate the quality of clustering results, and propose an efficient algorithm based on recursive bipartition to maximize an objective function of the proposed index. The optimization problem is NP-hard, so we approximate the semi-optimal solution via an implicitly restarted Lanczos method. One of the advantages of our algorithm is that the proposed index of each community in the clustering result is guaranteed to be higher than some predetermined threshold, $p$, which is completely controlled by users. We also account for the situation that $p$ is unknown. A statistical procedure of controlling both under-clustering and over-clustering errors simultaneously is carried out to select localized threshold for each subnetwork, such that the community detection accuracy is optimized. Accordingly, we propose a localized clustering algorithm based on binary tree structure. Finally, we exploit the stochastic blockmodels to conduct simulation studies and demonstrate the accuracy and efficiency of our algorithms, both numerically and graphically.

Study on Hesitant Fuzzy Information Measures and Their Clustering Application.

At present, research on hesitant fuzzy operations and measures is based on equal length processing, and an equal length processing method will inevitably destroy the original data structure and change the data information. This is an urgent problem to be solved in the development of hesitant fuzzy sets. Aiming at solving this problem, this paper firstly defines a hesitant fuzzy entropy function as the measure of the degree of uncertainty of hesitant fuzzy information and then proposes the concept of hesitant fuzzy information feature vector. The hesitant fuzzy distance measure and similarity measure are studied based on the information feature vector. Finally, the hesitant fuzzy network clustering method based on similarity measure is given, and the effectiveness of our algorithm through a numerical example is illustrated.

Energy-efficient clustering method for wireless sensor networks using modified gravitational search algorithm

Past decades have witnessed the advancement of wireless sensor networks (WSNs) in both academic and industrial communities. Clustering is one of the most popular methods to increase the lifespan of WSNs. The optimal number of cluster heads and how to organize the clusters are the most important issues to be addressed in the clustering methods. In this paper, we proposed a novel user-independent and dynamical method to calculate the optimal number of clusters, organize the clusters, and determine the best cluster heads in each round. In this method, efficient energy consumption and link quality were considered to compute the optimal number of clusters. Then, the algorithm began to organize the compact clusters with high energy level cluster heads. We investigated a new fitness function in order to achieve these objectives. A new version of gravitational search algorithm (GSA) was used to solve this optimization problem. In this algorithm, the power distance sums scaling method was applied to calculate the mass values. Then, a fuzzy logic controller is employed to identify the parameter of this algorithm to control the exploitation and exploration abilities of the method during the computational process of the algorithm. Then, the novel version of GSA was applied to reach an appropriate solution for the fitness function, find the optimal number of clusters, and properly organize these clusters. To evaluate the effectiveness of the proposed method, several experiments were performed and the obtained results were compared with the results of other popular clustering methods. The simulation results revealed that the performance of the modified GSA was better than other state-of-the-art meta-heuristic optimization algorithms. Moreover, the proposed method for the clustering problem in WSNs outperformed other popular clustering methods and increased the lifetime of WSNs.

Comparative Examination of Network Clustering Methods for Extracting Community Structures of a City From Public Transportation Smart Card Data

Human mobility data such as global positioning system (GPS) data from mobile phones and smart card data of public transportation have been used for the analysis of a city. Those studies have attempted to create networks whose nodes represent areas of the city and whose edges represent human flows between areas. Network clustering methods are applied to those networks for extracting community structures of cities. Although many studies have attempted to extract community structures of cities using human mobility data and network clustering methods, little has been explored about which method is the most effective, in terms of the accuracy and the amount of information of extracted clusters. In this paper, we propose evaluation metrics for evaluating network clustering methods based on the geographical cohesiveness, the regularity, and the amount of information of extracted clusters. We use smart card data of public transportation collected in Japan for testing the effectiveness of each method. We compare two types of origin-destination matrices for constructing networks and four types of network clustering methods including two types of bipartite network clustering methods. The results confirmed that a bipartite network clustering method considering multi-facet characteristics of each cluster and the functional relations between areas of a city shows the most accurate and richest information about community structures of cities.

Consensus Clustering of Tweet Networks via Semantic and Sentiment Similarity Estimation

Although Twitter has become an important source of information, the number of accessible tweets is too large for users to easily find their desired information. To overcome this difficulty, a method for tweet clustering is proposed in this paper. Inspired by the reports that network representation is useful for multimedia content analysis including clustering, a network-based approach is employed. Specifically, a consensus clustering method for tweet networks that represent relationships among the tweets' semantics and sentiment are newly derived. The proposed method integrates multiple clustering results obtained by applying successful clustering methods to the tweet networks. By integrating complementary clustering results obtained based on semantic and sentiment features, the accurate clustering of tweets becomes feasible. The contribution of this work can be found in the utilization of the features, which differs from existing network-based consensus clustering methods that target only the network structure. Experimental results for a real-world Twitter dataset, which includes 65 553 tweets of 25 datasets, verify the effectiveness of the proposed method.

A Two-Stage RPSO-ACS Based Protocol: A New Method for Sensor Network Clustering and Routing in Mobile Computing

In the field of mobile computing, after the placement of sensor nodes, energy efficiency of data transmission in the sensor network is becoming a critical issue as the sensor nodes are battery-operated devices. Most of the proposed protocols implement clustering based methods in order to balance the energy consumption among the sensor nodes through data aggregation. In this paper, a new method named a two-stage RPSO-ACS based protocol is proposed to improve the energy efficiency and network operation quality in the data transmission process, prolonging the lifespan of the network. Firstly, in the process of cluster head (CH) selection, the resampling particle swarm optimization (RPSO) is introduced, which shows better precision and efficiency than particle swarm optimization (PSO) and genetic algorithm (GA). Secondly, after the selection of cluster head, instead of the single-hop transmission from the cluster heads to base station (BS), which is used in many papers, we adopt multi-hop transmission to save energy. The ant colony optimization (ACO) is used to select the relay nodes between the cluster heads and the base station for the purpose of minimizing the energy consumption in the transmission process. In addition, we determine the reasonable threshold based on the energy consumption curve of the transmission. In this way, the nodes within a radius of the threshold from BS communicate with BS directly to save energy. Experimental results show that the proposed methods achieve the goal of improving the energy efficiency and operation quality of the sensor network in comparison to traditional methods.

Cloud Service Community Detection for Real-World Service Networks Based on Parallel Graph Computing

Heterogeneous information networks (e.g. cloud service relation networks and social networks), where multiple-typed objects are interconnected, can be structured by big graphs. A major challenge for clustering in such big graphs is the complex structures that can generate different results, carrying many diverse semantic meanings. In order to generate desired clustering, we propose a parallel clustering method for the heterogeneous information net-works on an efficient graph computation system (Spark). We use a multi-relation and path-based method to create similarity matrices, and implement our method based on graph computation model. It is inefficient to directly use existing data-parallel tools (e.g. Hadoop) for graph computation tasks, and some graph-parallel tools (e.g. Pregel) do not effectively address the challenges of graph construction and transformation. Therefore, we implemented our parallel method on the Spark system. The experiment results of clustering show our method is more accuracy.

Visualization and machine learning analysis of complex networks in hyperspherical space

A complex network is a condensed representation of the relational topological framework of a complex system. A main reason for the existence of such networks is the transmission of items through the entities of these complex systems. Here, we consider a communicability function that accounts for the routes through which items flow on networks. Such a function induces a natural embedding of a network in a Euclidean high-dimensional sphere. We use one of the geometric parameters of this embedding, namely the angle between the position vectors of the nodes in the hyperspheres, to extract structural information from networks. First we propose a simple method for visualizing networks by reducing the dimensionality of the communicability space to 3-dimensional spheres. Secondly, we use clustering analysis to cluster the nodes of the networks based on their similarities in terms of their capacity to successfully deliver information through the network. After testing these approaches in benchmark networks and compare them with the most used clustering methods in networks we analyze two real-world examples. In the first, consisting of a citation network, we discover citation groups that reflect the level of mathematics used in their publications. In the second, we discover groups of genes that coparticipate in human diseases, reporting a few genes that coparticipate in cancer and other diseases. Both examples emphasize the potential of the current methodology for the discovery of new patterns in relational data.

2-D quantization scheme utilizing SOFM neural network clustering for a DRoF system.

A two-dimensional (2-D) quantization scheme for the digitized radio-over-fiber (DRoF) transmission system is proposed and experimentally demonstrated. By using a self-organizing feature map (SOFM) neural network clustering method to implement the 2-D quantization, the spectral efficiency of the DRoF system can be effectively improved, and a better transmission performance is achieved. Compared with the conventional scalar quantization based on pulse code modulation in the common public radio interface, the sampled sequence is constructed as a 2-D array, and the SOFM neural network clustering method is employed to obtain an optimized quantization codebook, which can achieve a better error vector magnitude performance with the same quantization bits. In the experiments, the proposed scheme for the DRoF transmission is demonstrated in a 25km 5 Gbaud/λ four-level pulse amplitude modulation (PAM-4) intensity modulation and direct detection optical link. For the 5G NR standard, carrier aggregations of 27, 20, 16, and 13 100MHz orthogonal frequency division multiplexing with 4, 16, 64, and 256 QAM at quantization bits per sample of 3, 4, 5, and 6 are successfully achieved, respectively.

Regionalization of geographical space according to selected topographic factors in reference to spatial distribution of precipitation: application of artificial neural networks in GIS

In this paper, the regionalization of geographical space according to selected topographic factors and the spatial distribution of precipitation is discussed. The model takes into account qualitative and quantitative data describing the conditions associated with the studied precipitation. In the modelling, data mining methods including data clustering methods for agglomeration and artificial neural networks for classification have been used. The reason for their use was the classification of the area due to conditions related to precipitation, the distinguishing of similar areas and the delimitation of the propagation of the phenomenon or transition zones. To realize the research aims, professional software for data management, spatial data analysis, mathematical calculations and data mining have been used. The result of the research was a model of the classes representing areas with specific conditions affecting the phenomenon, transition zones between classes and areas with conditions other than those in the surroundings of the measuring stations, which are not classified in any of the classes. Classification results indicate the boundaries of the areas in which we can model the values measured at stations, the transition zones of possible discontinuous change and areas in which the phenomenon should not be modelled due to significantly different conditions from those in the neighbourhoods of measuring stations. Unclassified areas are also potential locations for new measuring stations.

Using Twitter network to detect market segments in the airline industry

The use of social media is increasingly considered a state-of-the-art practice for marketing campaigns. Researchers have investigated various social networks and media platforms to capture the behavior and characteristics of customers. This paper analyzes the official Twitter account of an airline company, Air New Zealand, to explore its market segments. To detect the communities of customers, we develop a network clustering method, which reveals the community classes of the network, along with a text mining analysis on each community detected by the cluster analysis. The results of the network analysis demonstrate that the social network of customers in the airline industry follows Pareto's principle that is similar to scale free networks. The findings of network clustering indicate Air New Zealand is essentially followed by New Zealand citizens. The local accounts are categorized into four communities: (1) lambda New Zealand citizens, (2) management, marketing, and digital media companies, (3) tourism and dining sectors, and (4) New Zealand sport players; while the global accounts fall into two communities: (1) worldwide celebrities and (2) the travel and aviation industry. The community detection method developed in this research is beneficial for marketing and customer strategy purposes as it enables airline companies to detect the categories of passengers interested into the brand. It also allows them to identify the potential sources for advertising by seeking out exceptionally connected customers who have high degrees of centrality.

Hospital-Nursing Home Transfer Patterns and Influence on Nursing Home Clostridium difficile Infection Rates

BackgroundLittle is known as to how hospital C. difficile infection (CDI) may impact nursing home (NH) CDI, or how patient transfers may modify this relationship. This study aims to examine a possible association between hospital and NH CDI rates, and whether NH CDI rates are influenced by patient transfers from hospital to NH.MethodsPatient transfers among the 5 hospitals and 34 NHs in Monroe County, NY were identified from the Minimum Data Set (MDS) 3.0 and Medicare Provider Analysis and Review files for 2011–13, and aggregated to the NH level. NH and hospital CDI rates were obtained from Emerging Infections Program CDI population surveillance and National Healthcare Safety Network data, respectively. Multivariate negative binomial regression modeled the association between hospital CDI rate (weighted by hospital-to-NH transfers/overall transfers among hospitals and NHs) and NH CDI rate, controlling for NH covariates from NH Compare and the Online Survey, Certification, and Reporting files. Patient transfer networks between hospitals and NHs were constructed, and basic network analysis of transfer patterns was conducted to confirm contributing factors to NH CDI rates from the multivariate model.ResultsWhen weighted hospital CDI rate increased by 1%, NH CDI rate increased by 18% (P = 0.016). Antibiotic and feeding tube prevalence were associated with a 4% and 8% increase in NH CDI rate, respectively (P≤0.014). Network analysis confirmed multivariate results and detected hospital-NH pairs with high edge weights (number of transfers) where NHs receiving patients from hospitals with high CDI rates had higher CDI rates. Network clustering methods were used to identify 2 sub-networks within overall annual networks and clusters of hospital-NH pairs for targeted intervention.ConclusionHospital CDI rate, adjusting for patient transfers, is associated with higher NH CDI rates in multivariate and network analyses, suggesting that NHs with a large inflow of patients from hospitals may need to implement stricter infection prevention practices to reduce transmission among residents. By identifying regional sub-networks, network analysis can also be used to actively manage facility CDI and prevent spread to other healthcare facilities.Disclosures All authors: No reported disclosures.

Impact of network clustering methods on pathway enrichment analysis tools

Impact of network clustering methods on pathway enrichment analysis tools

Нейросетевые методы кластеризации для оценки совокупности налогоплательщиков по степени их кредитоспособности

Нейросетевые методы кластеризации для оценки совокупности налогоплательщиков по степени их кредитоспособности

Self-Grouping Multi-Network Clustering.

Joint clustering of multiple networks has been shown to be more accurate than performing clustering on individual networks separately. Many multi-view and multi-domain network clustering methods have been developed for joint multi-network clustering. These methods typically assume there is a common clustering structure shared by all networks, and different networks can provide complementary information on this underlying clustering structure. However, this assumption is too strict to hold in many emerging real-life applications, where multiple networks have diverse data distributions. More popularly, the networks in consideration belong to different underlying groups. Only networks in the same underlying group share similar clustering structures. Better clustering performance can be achieved by considering such groups differently. As a result, an ideal method should be able to automatically detect network groups so that networks in the same group share a common clustering structure. To address this problem, we propose a novel method, ComClus, to simultaneously group and cluster multiple networks. ComClus treats node clusters as features of networks and uses them to differentiate different network groups. Network grouping and clustering are coupled and mutually enhanced during the learning process. Extensive experimental evaluation on a variety of synthetic and real datasets demonstrates the effectiveness of our method.

Tracing the Justice Conversation After “Green Cities, Growing Cities”

The article looks at various papers that cit the 1996 article Green cities, growing cities, just cities? Urban planning and the contradictions of sustainable development by S. Campbell. It offers a summary of the most-cited articles that originally cited Campbell including those written by professors Julian Agyeman, Philip Berke and Roger Keil. The author uses network clustering methods to group journal by how often these articles are cited.

Psychophysiological whole-brain network clustering based on connectivity dynamics analysis in naturalistic conditions.

We introduce a novel method for delineating context-dependent functional brain networks whose connectivity dynamics are synchronized with the occurrence of a specific psychophysiological process of interest. In this method of context-related network dynamics analysis (CRNDA), a continuous psychophysiological index serves as a reference for clustering the whole-brain into functional networks. We applied CRNDA to fMRI data recorded during the viewing of a sadness-inducing film clip. The method reliably demarcated networks in which temporal patterns of connectivity related to the time series of reported emotional intensity. Our work successfully replicated the link between network connectivity and emotion rating in an independent sample group for seven of the networks. The demarcated networks have clear common functional denominators. Three of these networks overlap with distinct empathy-related networks, previously identified in distinct sets of studies. The other networks are related to sensorimotor processing, language, attention, and working memory. The results indicate that CRNDA, a data-driven method for network clustering that is sensitive to transient connectivity patterns, can productively and reliably demarcate networks that follow psychologically meaningful processes. Hum Brain Mapp 37:4654-4672, 2016. © 2016 Wiley Periodicals, Inc.

KCI 등재 학술지의 분류를 위한 네트워크 군집화 방법의 비교

KCI 등재 학술지의 분류를 위한 네트워크 군집화 방법의 비교

NCMine: Core-peripheral based functional module detection using near-clique mining.

Motivation: The identification of functional modules from protein–protein interaction (PPI) networks is an important step toward understanding the biological features of PPI networks. The detection of functional modules in PPI networks is often performed by identifying internally densely connected subnetworks, and often produces modules with “core” and “peripheral” proteins. The core proteins are the ones having dense connections to each other in a module. The difference between core and peripheral proteins is important to understand the functional roles of proteins in modules, but there are few methods to explicitly elucidate the internal structure of functional modules at gene level.Results: We propose NCMine, which is a novel network clustering method and visualization tool for the core-peripheral structure of functional modules. It extracts near-complete subgraphs from networks based on a node-weighting scheme using degree centrality, and reports subgroups as functional modules. We implemented this method as a plugin of Cytoscape, which is widely used to visualize and analyze biological networks. The plugin allows users to extract functional modules from PPI networks and interactively filter modules of interest. We applied the method to human PPI networks, and found several examples with the core-peripheral structure of modules that may be related to cancer development.Availability and Implementation: The Cytoscape plugin and tutorial are available at Cytoscape AppStore. (http://apps.cytoscape.org/apps/ncmine).Contact: kengo@ecei.tohoku.ac.jpSupplementary information: Supplementary data are available at Bioinformatics online.

A dynamic semisupervised feedforward neural network clustering

AbstractAn efficient single-layer dynamic semisupervised feedforward neural network clustering method with one epoch training, data dimensionality reduction, and controlling noise data abilities is discussed to overcome the problems of high training time, low accuracy, and high memory complexity of clustering. Dynamically after the entrance of each new online input datum, the code book of nonrandom weights and other important information about online data as essentially important information are updated and stored in the memory. Consequently, the exclusive threshold of the data is calculated based on the essentially important information, and the data is clustered. Then, the network of clusters is updated. After learning, the model assigns a class label to the unlabeled data by considering a linear activation function and the exclusive threshold. Finally, the number of clusters and density of each cluster are updated. The accuracy of the proposed model is measured through the number of clusters, the quantity of correctly classified nodes, and F-measure. Briefly, in order to predict the survival time, the F-measure is 100% of the Iris, Musk2, Arcene, and Yeast data sets and 99.96% of the Spambase data set from the University of California at Irvine Machine Learning Repository; and the superior F-measure results in between 98.14% and 100% accuracies for the breast cancer data set from the University of Malaya Medical Center. We show that the proposed method is applicable in different areas, such as the prediction of the hydrate formation temperature with high accuracy.