Anupam Biswas

Investigating community structure in perspective of ego network

Defined Reachability and Isolability considering personalized view of ego network.Re-defined community structure giving more freedom to the nodes.Proposed ENBC algorithm for detecting communities.There is a trade-off between accuracy and quality metrics.ENBC is highly inclined towards the accuracy of communities. Complex relationships within the data are modeled as information network in various application areas of data mining. Identification of connected groups of nodes associated with similar information is one of the common interests behind such modeling. These connected groups are referred as community structure. This work investigates community structure by incorporating a sociological property known as ego network. Ego network facilitates personalized view of relationships. We introduce the notion of mutual interest in the relationship by utilizing such personalized view and re-define community structure. Different from classical way of defining communities through dense connectivity, proposed definition incorporates two properties: Reachability and Isolability. Reachability measures the ability of any node to reach out members of community, while Isolability accounts the ability of any community to isolate itself from rest of the network. Exploiting new definition of community structure, we propose an algorithm for identifying communities. Experimental results on a variety of real world data and synthetic data show communities identified with proposed algorithm is highly inclined towards accuracy in comparison to other state-of-the-art approaches.

An Empirical Study of Some Particle Swarm Optimizer Variants for Community Detection

Swarm based intelligent algorithms are widely used in applications of almost all domains of science and engineering. Ease and flexibility of these algorithms to fit in any application has attracted even more domains in recent years. Social computing being one such domain tries to incorporate these approaches for community detection in particular. We have proposed a method to use Particle Swarm Optimization (PSO) techniques to detect communities in social network based on common interest of individual in the network. We have performed rigorous study of four PSO variants with our approach on real data sets. We found orthogonal learning approach results quality solutions but takes reasonable computation time on all the data sets for detecting communities. Cognitive avoidance approach shows average quality solutions but interestingly takes very less computation time in contrast to orthogonal learning approach. Linear time varying approach performs poorly on both cases, while linearly varying weight along with acceleration coefficients is competitive to cognitive avoidance approach.

Defining quality metrics for graph clustering evaluation

Proposed a set of three quality metrics AVI, AVU and ANUI.Roughly four kind of analysis have been performed.AVI, AVU and ANUI together can give better indication about accuracy.These metrics satisfy all of the six quality related properties.Linearity in characteristics of our metrics can also give indication about accuracy. Evaluation of clustering has significant importance in various applications of expert and intelligent systems. Clusters are evaluated in terms of quality and accuracy. Measuring quality is a unsupervised approach that completely depends on edges, whereas measuring accuracy is a supervised approach that measures similarity between the real clustering and the predicted clustering. Accuracy cannot be measured for most of the real-world networks since real clustering is unavailable. Thus, it will be advantageous from the viewpoint of expert systems to develop a quality metric that can assure certain level of accuracy along with the quality of clustering.In this paper we have proposed a set of three quality metrics for graph clustering that have the ability to ensure accuracy along with the quality. The effectiveness of the metrics has been evaluated on benchmark graphs as well as on real-world networks and compared with existing metrics. Results indicate competency of the suggested metrics while dealing with accuracy, which will definitely improve the decision-making in expert and intelligent systems. We have also shown that our metrics satisfy all of the six quality-related properties.

Analyzing evolutionary optimization and community detection algorithms using regression line dominance

Developed regression line dominance and shifting mechanism.Proposed visual analysis method for evolutionary optimization algorithms.Community detection algorithms are also analyzed with proposed method.Developed one-to-one, one-to-many and many-to-many comparison methodology.Experimented with 25 benchmark functions and 10 real-world networks. In this paper, a visual analysis methodology is proposed to perform comparative analysis of guided random algorithms such as evolutionary optimization algorithms and community detection algorithms. Proposed methodology is designed based on quantile-quantile plot and regression analysis to compare performance of one algorithm over other algorithms. The methodology is extrapolated as one-to-one comparison, one-to-many comparison and many-to-many comparison of solution quality and convergence rate. Most of the existing approaches utilize both solution quality and convergence rate to perform comparative analysis. However, the many-to-many comparison i.e. ranking of algorithms is done only with solution quality. On the contrary, with proposed methodology ranking of algorithms is done in terms of both solution quality and convergence rate. Proposed methodology is studied with four evolutionary optimization algorithms on 25 benchmark functions. A non-parametric statistical analysis called Wilcoxon signed-rank test is also performed to verify the indication of proposed methodology. Moreover, methodology is also applied to analyze four state-of-the-art community detection algorithms on 10 real-world networks.

Community-based link prediction

This work proposes a community-based link prediction approach for identifying missing links or the links that are likely to appear in near future. Earlier works on link prediction consider only connectivity pattern or node attributes. We incorporate the notion of community structure in link prediction. An algorithm is designed to account the influence of communities on link prediction. We have considered recently developed edge centrality measures to compute likelihood scores of missing links. The performance of proposed algorithm is analyzed in terms of three metrics and execution time on both real-world networks and synthetic networks, where ground truth communities are already defined. The time complexity of proposed algorithm is also analyzed.

Swarm Intelligence Techniques and Their Adaptive Nature with Applications

Swarm based techniques have huge application domain covering multiple disciplines, which include power system, fuzzy system, forecasting, bio-medicine, sociological analysis, image processing, sound processing, signal processing, data analysis, process modeling, process controlling etc. In last two decades numerous techniques and their variations have been developed. Despite many variations are being carried out, main skeleton of these techniques remain same. With diverse application domains, most of these techniques have been modified to fit into a particular application. These changes undergo mostly in perspective of encoding scheme, parameter tuning and search strategy. Sources of real world problems are different, but their nature sometimes found similar to other problems. Hence, swarm based techniques utilized for one of these problems can be applied to others as well. As sources of these problems are different, applicability of such techniques are very much dependent on the problem. Same encoding scheme may not be suitable for the other similar kind of problems, which has led to development of problem specific encoding schemes. Sometimes found that, even though encoding scheme is compatible to a problem, parameters used in the technique does not utilized in favor of the problem. So, parameter tuning approaches are incorporated into the swarm based techniques. Similarly, search strategy utilized in swarm based techniques are also vary with the application domain. In this chapter we will study those problem specific adaptive nature of swarm based techniques. Essence of this study is to find pros and cons of such adaptation. Our study also aims to draw some aspects of such problem specific variations through which it can be predicted that what kind of adaptation is more convenient for any real world problem.

Visual Analysis of Evolutionary Optimization Algorithms

Growing application of evolutionary optimization algorithms in the problems of different domain have led to analyze their efficiency and effectiveness rigorously. Various approaches have been proposed to algorithms for performance evaluation such as parametric, non-parametric or mathematical which lack direct involvement of results obtained. A visual comparative performance evaluation method has been proposed in this paper incorporating more direct participation of results. Proposed method has been studied in perspective of three types of comparisons one-to-one, one-to-many and many-to-many. Necessary interpretations for the method have been illustrated and examined with solutions obtained on several benchmark functions through well known evolutionary optimization algorithms.

Regression line shifting mechanism for analyzing evolutionary optimization algorithms

This work introduces a novel methodology to perform the comparative analysis of evolutionary optimization algorithms. The methodology relies simply on linear regression and quantile–quantile plots. The methodology is extrapolated as the one-to-one comparison, one-to-many comparison and many-to-many comparison of solution quality and convergence rate. Most of the existing approaches utilize both solution quality and convergence rate to perform comparative analysis. However, many-to-many comparison, i.e., ranking of algorithms is done only in terms of solution quality. The proposed method is capable of ranking algorithms in terms of both solution quality and convergence rate. Method is analyzed with well-established algorithms and real data obtained from 25 benchmark functions.

An Atomic Model Based Optimization Algorithm

In this paper, an optimization algorithm is proposed for solving combinatorial optimization problems. The designing of the algorithm is based on the Bohrs atomic model. Microcosm of physics and chemistry such as excitation and de-excitation of electrons in atom and atoms bond formation phenomenon are the key mechanism of this algorithm. Atoms form compounds to minimize the energy of their electrons and become more stable, with this quest positively charged atoms attracts negatively charged atoms to form bonds among them. Utilizing these phenomenon to attain more stable state of an atom, a novel strategy is incorporated into the algorithm. This stable state is resembled with global optima in the proposed algorithm. Performance of proposed algorithm is analyzed on real life combinatorial optimization problems.

Agglomerative Approaches for Partitioning of Networks in Big Data Scenarios

Big Data systems are often confronted with storage and processing-related issues. Nowadays, data in various domains is growing so enormously and so quickly that storage and processing are becoming the two key concerns in such large systems of data. In addition to the size, complex relationship within the data is making the system highly sophisticated. Such complex relationships are often represented as network of data objects. Parallel processing, external memory algorithms, and data partitioning are at the forefront of techniques to deal with the Big Data issues. This chapter discusses these techniques in relation to storage and processing of Big Data. The Big Data partitioning techniques, such as agglomerative approaches in particular, have been studied and reported. Network data partitioning or clustering is common to most of the network-related applications where the objective is to group similar objects based on the connectivity among them. Application areas include social network analysis, World Wide Web, image processing, biological networks, supply chain networks, and many others. In this chapter, we discuss the relevant agglomerative approaches. Relative advantages with respect to Big Data scenarios are also presented. The discussion also covers the impact on Big Data scenarios with respect to strategic changes in the presented agglomerative approaches. Tuning of various parameters of agglomerative approaches is also addressed in this chapter.