Search

Accumulative structure or cluster-like shape is one of the important features of social networks. These structures and clusters are communities in a complex network and are fully detectable. Common group behaviors of different communities can be categorized using community detection methods. Categorize behavior allows the study of each part of the network to be done centrally. This paper uses trust-based centrality to detect the communities that make up the network. Centrality determines the relative importance of a node in the graph of social networks. Redefining the trust-based centrality makes it possible to change the position in the analysis of centrality and separates the local central nodes and global central nodes. Then, a trust-based algorithm is proposed to express the strength of trust penetration conceptually between nodes to extract communities in networks. This method has led to the achievement of a flexible and effective community detection method. The proposed algorithm is applied to four benchmark networks. The experiments consist of two independent parts. The first part is to use the proposed algorithm to detect clusters and communities. After that, the algorithm is compared with a Girvan–Newman inspired method. The second part is the implementation of the proposed algorithm with a large number of iterations with the aim of modularity maximization and comparing it with other community detection algorithms. Although, the modularity criterion has been used to validate and compare the solution quality in both independent parts of the experiments. The results show about 1.4–5.2% improvement in community detection.

Aromatic compounds are organic matters that have the form of carbon rings, such as benzene rings. The compound is contained in most real-life chemicals, including medicines, detergents, and cosmetics. Because mutagenicity of these substances can pose a great health risk if they enter the human body, predicting mutagenicity and avoiding risk are important problems. Due to recent advances in deep learning, many studies have been conducted to predict the mutagenicity of graphically expressed aromatic compounds, and they are performing well. However, previous methods that used deep learning to predict the mutagenicity of molecules lead to the problem of dilution of local information in molecular structures. In graph neural networks, the embedding of graphs is determined by the average or sum of node embeddings, and when a particular node dominates, information from local nodes is diluted. In this paper, we propose a model that learns molecules’ local information properties to be undiluted by separating them from the original graph. By partitioning the graph, we preserve local information by breaking the relationship between the dominant nodes and the non-dominant nodes so that they do not affect each other's information updates. Experiments show that we successfully partition carbon rings and functional groups in molecular graphs using the Girvan Newman algorithm, and embed the segmented graphs using the same neural networks, allowing neural networks to learn about carbon rings and functional groups. Comparisons with other neural networks confirm performance improvements on 1% accuracy.

Community discovery is an important research content in complex network analysis and an important tool for community analysis. It divides nodes into clusters according to the similarity of nodes or similarities in structure. This paper takes the EU public procurement contract network from 2008 to 2016 as an example, selects 6 countries with different data sizes, Austria, Cyprus, Germany, France, Hungary and Italy, analyzes the characteristics of the contract network and evaluates different community classification algorithms Performance. Taking the running speed and modularity as the evaluation criteria, comparing the Louvain algorithm and Girvan Newman algorithm the speed and accuracy of these 2 methods on different scale data sets, analyze the most suitable contract network community classification algorithm. Through analysis, the contract network is a sparse bipartite network, Louvain algorithm is a more suitable community detection method than Girvan Newman algorithm for contract networks. It has a fast running speed and high accuracy.

Analiza društvenih mreža postala je vrlo dinamično i unosno područje pojavom online društvenih mreža i povećanjem kapaciteta za njihovu obradu. Otkrivanje zajednica jedno je od najvažnijih područja unutar analize društvenih mreža. U ovom radu, predstavljen je pojam društvene mreže, njene karakteristike, pojam zajednice u kontekstu društvenih mreža i problematika procjene njihove kvalitete. Nadalje, napravljena je usporedba Girvan-Newman, Label Propagation i Clauset-Newman-Moore algoritama za otkrivanje zajednica. Analiza je provedena u dva dijela: prvo na umjetno generiranim mrežama proizvedenih Lancichinetti-Fortunato-Radicchi metodom, a potom na stvarnim podacima s društvene mreže Twitter. The arrival of online social networks and the increase in the capacity to process them has caused social network analysis to become a very dynamic and lucrative field. Community detection is one of the most important areas of social network analysis. This thesis presents the notion of a social network, its characteristics, the notion of community in the context of social networks and the problem of assessing their quality. Furthermore, a comparison of Girvan-Newman, Label Propagation and Clauset-Newman-Moore community detection algorithms was made. The analysis was performed in two parts: first on artificially generated networks produced by the Lancichinetti-Fortunato-Radicchi benchmark, and then on real data from the social media network Twitter.

The term community defines and explains the fact that there exists a group where the members of that group happen to have more commonality amongst each other in terms of features, comparing to the feature of that two members of different groups which would have fewer things in common. In a world that is ruled by online social networking, it is important and difficult to understand that based on what opinions or environment leads to the formation of a certain community amongst the masses. These views tends to become the foundation of society and hence country or even the world. In this paper, we attempt to understand this by experimenting on a popular social network for video sharing, i.e., YouTube. Our work consists of community detection on a real-time dataset of related videos about Indian political news. We try to get the best possible structure by choosing a suitable algorithm depending upon the scenario of our platform. As a result, the post-detection of communities, we extracted as many features possible within each group. The final step is content analysis to obtain the knowledge that we infer from each community.

Community detection is an important tool for analyzing networks; it can help us understand the structures and functional characteristics. Network communities represent a principled way of organizing real-world networks into densely connected groups of nodes, whereas a community is a cluster of nodes that are strongly connected to each other in a subnetwork than to the rest of the network. This has remarkable results in various fields, e.g., social science, bibliometrics, marketing and recommendations, biology etc. This research proposes a dataset based on COVID19 distribution. The network dataset is formed by tracing the transmission of the virus among the world countries. This experiment demonstrates the spread of COVID19 and its mutated strain. Several algorithms such as Girvan Newman, Greedy, Louvain, Clustering, and Label Propagation are implemented on the proposed dataset, and the results are evaluated using the modularity measure. The proposed COVID 19 dataset demonstrated the properties and dynamic behavior similar to Zachry’s Karate dataset.

Community detection in the complex network is the process of finding optimal clusters of vertices that are similar in characteristics. To study the properties and functions of complex networks, community detection plays a crucial role. Community detection is generally categorized as an optimization problem, and due to its inherent property it can’t be solved by the traditional optimization method, and over the past few decades, various algorithms have been proposed to address this problem in multiple fields such as power system, physics, biology, or sociology. In this paper, we present a critical survey on various algorithms for community detection currently available such as genetic algorithm, evolutionary algorithms, a nature-inspired algorithm, deep learning algorithm. This survey paper outlines the challenges and constraints of different state-of-arts community algorithms detection by utilizing contemporary techniques like deep neural networks, genetic algorithms, and various topological features based methodologies. This paper includes two genetic algorithms one is based in safe and balanced initialization while the other is based in clustering co-efficient and common finding for these two algorithms are they both can be further improved for the detection of overlapping communities and also we discussed community detection using internal force between vertices this can method can further include the dynamic strategies to uncover the communities so that it will work for the dynamic network and increase the practicality of the method this paper also include community detection method using edges betweenness centrality this method can further be used improved in the detection of directed and weighted networks lastly community detection using deep neural network auto-encoder is used to for dimension reduction and further K-means is used for the detection of community in the network this method can be further improved in term of scaling of the model and can be further test be tested ob big dataset which thousands of vertices and edges.

As a typical representative of research in various scientific domains, methods and models always play as a critical role. These methods and models are numerous and heterogeneous, and usually need to be manually sorted out and classified systematically, which is often time-consuming and laborious. Therefore, this paper proposes a solution for methods and models classification automatically based on community detection algorithms in the scientific domain. The co-words network extracted from the literature data could be divided into different communities based on different detection algorithms. Meanwhile, the keywords of methods and models are automatically divided into different categories. As a case study, this paper analyzed the literature data in the human resource management (HRM) domain. We compared the results obtained by six community detection algorithms, such as Louvain, Girvan Newman (GN), with technical indicators and expert validation. At last, the keywords of the methods and models in the HRM domain classified into 12 communities. The results shown that our research provides an effective method that can effectively classify the methods and models automatically in the scientific domain.

Complex network theory is introduced to solve the islanding problem in an emergency of distribution networks. In this study, the authors put forward an intentional islanding method based on community detection. In this method, a new index has been defined called electrical edge betweenness, on the strength of edge betweenness in complex networks, which fuses electrical characteristics with topological features of actual power lines. Based on the index, the Girvan–Newman algorithm is employed to detect the community structure of distribution networks. Through referring to the modularity value (function Q) and coherent generator groups, they can get a reasonable amount and regions of communities. Then the whole distribution network can be partitioned into several self-sustainable islands meeting the stable operation constraints. The effectiveness of the authors’ proposed method is tested on a standard IEEE 118-bus system.

Betweenness centrality quantifies the importance of nodes in a graph in many applications, including network analysis, community detection and identification of influential users. Typically, graphs in such applications evolve over time. Thus, the computation of betweenness centrality should be performed incrementally. This is challenging because updating even a single edge may trigger the computation of all-pairs shortest paths in the entire graph. Existing approaches cannot scale to large graphs: they either require excessive memory (i.e., quadratic to the size of the input graph) or perform unnecessary computations rendering them prohibitively slow. We propose $i$ Central ; a novel incremental algorithm for computing betweenness centrality in evolving graphs. We decompose the graph into biconnected components and prove that processing can be localized within the affected components. $i$ Central is the first algorithm to support incremental betweeness centrality computation within a graph component. This is done efficiently, in linear space; consequently, $i$ Central scales to large graphs. We demonstrate with real datasets that the serial implementation of $i$ Central is up to 3.7 times faster than existing serial methods. Our parallel implementation that scales to large graphs, is an order of magnitude faster than the state-of-the-art parallel algorithm, while using an order of magnitude less computational resources.

China has extensively implemented the Joint Prevention and Control of Atmospheric Pollution (JPCAP) strategy to address regional air pollution challenges; however, the lack of accurate scope and regional priority control sequence has resulted in reduced pollution control efficiency and high air quality improvement costs. We proposed the following new solutions to these problems: (i) constructing inter-regional network correlation models based on three key network topology parameters and (ii) subdividing large regions into sub-regions using the Girvan Newman (GN) algorithm. To test these methods, we applied them to a case study of China's provincial-level JPCAP areas using hourly air resource endowment (ARE) data obtained from 2013 to 2017. We found that the modularity (Q) values of ARE network partitioning in different seasons were approximately 0.5, indicating the well-structured association of the network constructed in this study. Seven areas were identified based on annual values, while eight, five, four, and six areas were classified for spring, summer, autumn, and winter, respectively. The strength of the interconnectedness of the provinces in the region, as measured by the clustering coefficient ( C ˜ ) value, is ordered as follows: Autumn > winter > spring > summer. For a specific JPCAP region, the priority control level can be dynamically adjusted based on the Pearson correlation coefficient (r-value) between the target city or province and the surrounding region. The regional division results, priority control sequences, and key control seasons identified in this study can provide technical and methodological reference for delineating and optimizing China's JPCAP policy.

A Self-Contained Girvan-Newman Algorithm and Mean Variance Model combined approach is proposed to allocate the defense resource in road dangerous goods transportation network. Firstly, the weighted physical network without direction and its weighted service network with direction are established. The Self-Contained Girvan-Newman Algorithm is applied to separate the whole weighted physical network into several communities. Next, based on the service network, the covariance matrix of each separated community is established, the Mean Variance Model is used to allocate the defense resource for each community, which focuses on selecting the option with the lowest probability of loss caused by the dangerous goods transportation accident/risk. The case study is conducted by using the road network and the dangerous goods transportation volume of Dalian, China as the background. When the whole network is separated into 6 communities, the defense resource capability in the whole network is the best. The Power Function Allocation Model (PFAM) is applied as the comparison approach, the overall rescue capability of the network is defined and applied to evaluate the defense resource allocation schemes. The results show that, the approach proposed in this paper has better effectiveness than PFAM, especially when the whole network is separated into 6 communities.

This research used a cell structure map to visualize technological evolution and showed the developmental trend in a technological field. The basic concept was to organize patents into a map produced by growing cell structures. The map was then disassembled into clusters with similar contexts using the Girvan---Newman algorithm. Next, the continuity between clusters in two snapshots was identified and used as the base for establishing a trajectory in the technology. An analysis of patents in the flaw detection field found that the field was composed of several technological trajectories. Among them, ultrasonic flaw detection, wafer inspection and substrate inspection were relatively larger and more continuing technologies, while infrared thermography defect inspection has been an emerging topic in recent years. It is to be hoped that the map of technology constructed in this research provides insights into the history of technological evolution and helps explain the transition patterns through changes in cluster continuity. This can serve a reference point by experts who attempt to visualize the mapping of technological development or identify the latest focus of attention.

In the current era, the analysis of social network data is one of the challenging tasks. Social networks are represented as a graphical structure where the users will be treated as nodes and the edges represent the social tie between the users. Research such as community identification, detection of centrality, detection of fraud, prediction of links and many other social issues are carried out in social network analysis. However, the community plays an important role in solving major issues in a real-world scenario. In a community structure, nodes inside a community are densely connected, whereas nodes between the communities are sparsely connected. Determining the emotional behavior from communities is the major concern because, emotional behavior of community helps us to solve problems like brand reachability, find target audience, build brand awareness and much more. Ekman’s emotional scale is a popular categorical model which assumes that there is a finite number of basic and discrete emotions and is used to classify the emotions. In this paper, a novel method is proposed to determine the community-specific emotional behavior of the users related to a particular topic. Communities are formed based on network topology rather than emotions. Girvan Newman algorithm is used to construct the communities of different users, who share their views on twitter media platforms regarding a topic. Then Ekman’s emotional scale is used to categorize the emotions of the users of each community. This identifies how the people of different communities react to an incident. The incident can be treated as NEWS/Trending threads on Twitter/Facebook shares. The emotional analysis is done community-specific so that the behavioral analysis of an incident is performed specifically for that community. Further, the comprehensive experimental analysis shows that the proposed methodology constructs influential communities and performs emotional analysis efficiently.

U ovom završnom radu obrađen je jedan od najvažnijih problema u analizi društvenih mreža, a to je otkrivanje zajednica. Opisana je struktura društvenih mreža i mjere pomoću kojih se one analiziraju. Objašnjeni su postupci grupiranja u teoriji grafova. Proučeni su i opisani Girvan-Newman i Clauset-Newman algoritmi za otkrivanje nepreklapajućih zajednica u društvenim mrežama. Navedeni algoritmi provedeni su nad stvarnim skupom podataka te nad umjetno stvorenim skupom primjenom Watts-Strogatz modela te je dana usporedba njihovih rezultata. Opisan je pojam small-world fenomena te razvoj i svojstva Watts-Strogatz modela. This paper discusses one of the most important issuses in social network analysis which is community detection. A detailed description of a social network structure is given and measures used to analyze them. Different clustering techniques used in graph theory are also described. Steps of Girvan-Newman and ClausetNewman Moore algorithms for community detection are shown. The two algorithms are implemented in Snap.py library and evaluated on five real-world datasets as well as on generated random small-world graphs. Small-world phenomenon is described along with the evolution and characteristics of the Watts-Strogatz model.

The discovery and analysis of community structure in complex networks is a hot issue in recent years. In this paper, based on the fast greedy clustering algorithm CNM with the thought of local search, the introduction of the idea of node centrality and the optimal division of the central nodes and their neighbor nodes into correct communities, a new algorithm CNM-Centrality of detecting communities in complex networks is proposed. In order to verify the accuracy and efficiency of this algorithm, the performance of this algorithm is tested on several representative real-world networks and a set of computer-generated networks by LFR-benchmark. The experimental results indicate that this algorithm can identify the communities accurately and efficiently. Furthermore, this algorithm can also acquire higher values of modularity and NMI than the CNM, Infomap, Walktrap algorithms do.

Community detection is a critical task for complex network analysis. It helps us to understand the properties of the system that a complex network represents and has significance to a wide range of applications. Nowadays, the challenges faced by community detection algorithms include overlapping community structure detection, large scale network analysis, dynamic changing of analyzed network topology and many more. In this paper a self-organizing community detection algorithm, based on the idea of swarm intelligence, was proposed and its parallel algorithm was designed on Giraph++ which is a semi-asynchronous parallel graph computation framework running on distributed environment. In the algorithm, a network of large size is firstly divided into a number of small sub-networks. Then, each sub-network is modeled as a self-evolving swarm intelligence sub-system, while each vertex within the sub-network acts iteratively to join into or leave from communities based on a set of predefined vertex action rules. Meanwhile, the local communities of a sub-network are sent to other sub-networks to make their members have a chance to join into, therefore connecting these self-evolving swarm intelligence sub-systems together as a whole, large and evolving, system. The vertex actions during evolution of a sub-network are sent as well to keep multiple community replicas being consistent. Thus network communication efficiency has a great impact on the algorithm' performance. While there is no vertex changing in its belonging communities anymore, an optimal community structure of the whole network will have emerged as a result. In the algorithm it is natural that a vertex can join into multiple communities simultaneously, thus can be used for overlapping community detection. The algorithm deals with vertex and edge adding or deleting in the same way as the algorithm running, therefore inherently supports dynamic network analysis. The algorithm can be used for the analysis of large scale networks with its parallel version running on distributed environment. A variety of experiments conducted on synthesized networks have shown that the proposed algorithm can effectively detect community structures and its performance is much better than certain popular community detection algorithms.

In this paper, we present a novel data clustering framework for big sensory data produced by IoT applications. Based on a network representation of the relations among multi-dimensional data, data clustering is mapped to node clustering over the produced data graphs. To address the potential very large scale of such datasets/graphs that test the limits of state-of-the-art approaches, we map the problem of data clustering to a community detection one over the corresponding data graphs. Specifically, we propose a novel computational approach for enhancing the traditional Girvan–Newman (GN) community detection algorithm via hyperbolic network embedding. The data dependency graph is embedded in the hyperbolic space via Rigel embedding, allowing more efficient computation of edge-betweenness centrality needed in the GN algorithm. This allows for more efficient clustering of the nodes of the data graph in terms of modularity, without sacrificing considerable accuracy. In order to study the operation of our approach with respect to enhancing GN community detection, we employ various representative types of artificial complex networks, such as scale-free, small-world and random geometric topologies, and frequently-employed benchmark datasets for demonstrating its efficacy in terms of data clustering via community detection. Furthermore, we provide a proof-of-concept evaluation by applying the proposed framework over multi-dimensional datasets obtained from an operational smart-city/building IoT infrastructure provided by the Federated Interoperable Semantic IoT/cloud Testbeds and Applications (FIESTA-IoT) testbed federation. It is shown that the proposed framework can be indeed used for community detection/data clustering and exploited in various other IoT applications, such as performing more energy-efficient smart-city/building sensing.

Community Detection is starting to become an inevitable task for data science community and like-minded analysts to understand the hidden insights and trends in large complex networks. Through this paper we would like to understand and comparatively analyze the prevalent and recent community detection techniques like Minimum Cut Method, Hierarchical Clustering, Girvan Newman Algorithm, Modularity Maximization, Statistical Inference, Clique Percolation & Maximum Permanence Method based on their network relevance, adaptability and robustness to changes and finally their capacity of size and speed. Finally, we will review the realworld application of these algorithms.

Most of the real-world networks exhibit community structure, a property that reveals the existence of natural vertex clusters whose inter-edge density is lower than intra-edge density between various groups. Despite providing a better understanding of network structure and characteristics, community detection has many practical applications in diverse domains. Communities obtained from the telephone network provides many useful information that can be used for churn prediction, budget control in organizations etc. Detecting communities is a fundamental need in the area of networks, yet challenging. In this paper, we propose an extension to the Girvan-Newman algorithm for finding the betweenness using the transitive closure property and the greedy technique in Dijkstra's single source shortest path method.

Devise an algorithm for updating betweenness centrality in fully dynamic graphs.Recalculate centrality without computing all pairs shortest paths in the entire graph.Adapt a highest centrality edge finding algorithm based on the proposed algorithm.Adapt a community detection algorithm using the proposed algorithms.Experimental results show the proposed algorithm outperforms existing algorithms. Betweenness centrality of a vertex (edge) in a graph is a measure for the relative participation of the vertex (edge) in the shortest paths in the graph. Betweenness centrality is widely used in various areas such as biology, transportation, and social networks. In this paper, we study the update problem of betweenness centrality in fully dynamic graphs. The proposed update algorithm substantially reduces the number of shortest paths which should be re-computed when a graph is changed. In addition, we adapt a community detection algorithm using the proposed algorithm to show how much benefit can be obtained from the proposed algorithm in a practical application. Experimental results on real graphs show that the proposed algorithm efficiently update betweenness centrality and detect communities in a graph.

In this paper, we propose an algorithm to find subgraphs with given properties in large social networks. A computational experiment that confirms the effectiveness of the proposed algorithm is presented.

Understanding and detecting natural communities in networks have been a fundamental challenge in networks, and in science generally. Recently, we proposed a hypothesis that homophyly/kinship is the principle of natural communities based on real network experiments, proposed a model of networks to explore the principle of natural selection in nature evolving, and proposed the measure of structure entropy of networks. Here we proposed a community finding algorithm by our measure of structure entropy of networks. We found that our community finding algorithm exactly identifies almost all natural communities of networks generated by natural selection, if any, and that the algorithm exactly identifies or precisely approximates almost all the communities planted in the networks of the existing models. We verified that our algorithm identifies or very well approximates the ground-truth communities of some real world networks, if the ground-truth communities are semantically well-defined, that our algorithm naturally finds the balanced communities, and that the communities found by our algorithm may have larger modularity than that by the algorithms based on modularity, for some networks. Our algorithm provides for the first time an approach to detecting and analyzing natural or true communities in real world networks. Our results demonstrate that structure entropy minimization is the principle of detecting the natural or true communities in large-scale networks.

Community detection in the social networks is one of the most important tasks of social computing. Highly relevant researches indicate that the social network generally contains both an overlapping and hierarchical structure. This paper introduces an efficient and functional community detection algorithm MOHCC, which can concurrently discover overlapping and hierarchical organization in complex networks. This algorithm first extracts all maximal cliques from the original complex network. Merges all extracted maximal cliques into a dendrogram by using the aggregative framework presented in MOHCC. Finally, it cuts through the dendrogram and obtains a network partition with maximum extended partition density. Experimental results utilizing computer-generated artificial networks and real-world social benchmark networks give satisfactory correspondence.

Betweenness centrality is a centrality measure that is widely used, with applications across several disciplines. It is a measure which quantifies the importance of a vertex based on its occurrence in shortest paths between all possible pairs of vertices in a graph. This is a global measure, and in order to find the betweenness centrality of a node, one is supposed to have complete information about the graph. Most of the algorithms that are used to find betwenness centrality assume the constancy of the graph and are not efficient for dynamic networks. We propose a technique to update betweenness centrality of a graph when nodes are added or deleted. Our algorithm experimentally speeds up the calculation of betweenness centrality (after updation) from 7 to 412 times, for real graphs, in comparison to the currently best known technique to find betweenness centrality.

The centerpiece of biodiversity conservation is ensuring cooperation among actors involved in the management of protected areas. To contribute to the debate in the field of protected areas governance, we investigate the characteristics of two ego networks established in two Natura 2000 protected areas in Romania. After evaluating the most common ego-network metrics that demonstrate the characteristics of each network, we analyzed the strong ties by using Simmelian ties within protected areas management ego-networks and clustered the embedded links in three primary Girvan Newman groups. The study provides insights regarding several means to improve the cooperation of environmental conservation in Natura 2000 areas.

Social network is one of the vibrant topic and allow researcher to think in various dimension to provide results with wide scope. In this paper we are aimed at optimizing cluster to enhance the result. It helps to increase the computational performance of a social network as well as maintaining the valid nodes for bidirectional interaction between end users. Proposed methodology applies Girvan Newman clustering algorithm along with minimum cut algorithm for optimization purpose which makes effective interaction with in social communication network. Our research includes two major phases in it. During first phase we apply Girvan Newman algorithm to cluster the nodes with common behavior, properties and characteristics. In the second phase, application of minimum cut algorithm will be done into clustered social network nodes. The major objective is to analyze network nodes and make them stable and active and to remove the nodes that are not interacting with any network nodes that is inactive nodes. Such inactive nodes make the less efficiency and inaccurate communication. For these kind of problem, we are going to apply minimum cut algorithm. It will maximize the effective bidirectional interaction by removing the unwanted, noisy and inconsistent nodes from these clusters.

The determination of community structures within social networks is a significant problem in the area of data mining. A proper community is usually defined as a subgraph with a higher internal density and a lower crossing density with others subgraphs. Hierarchical clustering algorithms produce a set of nested clusters, sometimes called dense subgraphs, organized as a hierarchical system and the output is always referred as a dendrogram. However, determining which of clusters in the dendrogram will be selected to form communities in the final output is a difficult problem. Most implementations of data mining algorithms require expert guidance in the implementation of the algorithm in order to establish the appropriate selection of such communities, and ultimately the output may not be optimized as with fixed height tree-cutting algorithms. In this paper, a novel algorithm for community selection is proposed. The intuition of our approach is based on drops of densities between each pair of parent and child nodes on the dendrogram - the higher the drop in density, the higher probability the child should form an independent community. Based on the Max-Flow Min-Cut theorem, we propose a novel algorithm which can output an optimal set of local communities automatically. In addition, a faster algorithm running in linear time is also presented for the case that the dendrogram is a tree. Finally, we validate this approach through a variety of data sets ranging from synthetic graphs to real world benchmark data sets.