Proposed Clustering Scheme Research Articles

Network Lifetime Maximization Clustering Scheme in IEEE 802.15.4 Networks

An energy-efficient clustering scheme to maximize the network lifetime is presented in IEEE 802.15.4 networks. In the proposed clustering scheme, even though the cluster is divided into several sub-clusters in order to decrease data redundancies, the sub-CH does not transmit the beacon frame due to the problem of beacon collision. Our clustering scheme also allows the CH to control the size of the sub-cluster according to the residual energy of the sub-CH. The performance of the proposed scheme is verified by simulations that demonstrate how our scheme provides a better network lifetime than the conventional scheme.

Spectral Preprocessing for Clustering Time-Series Gene Expressions

Based on gene expression profiles, genes can be partitioned into clusters, which might be associated with biological processes or functions, for example, cell cycle, circadian rhythm, and so forth. This paper proposes a novel clustering preprocessing strategy which combines clustering with spectral estimation techniques so that the time information present in time series gene expressions is fully exploited. By comparing the clustering results with a set of biologically annotated yeast cell-cycle genes, the proposed clustering strategy is corroborated to yield significantly different clusters from those created by the traditional expression-based schemes. The proposed technique is especially helpful in grouping genes participating in time-regulated processes.

A distributed group mobility adaptive clustering algorithm for mobile ad hoc networks

This paper proposes a distributed group mobility adaptive (DGMA) clustering algorithm for mobile ad hoc networks (MANETs) on the basis of a revised group mobility metric, linear distance based spatial dependency ( LDSD), which is derived from the linear distance of a node’s movement instead of its instantaneous speed and direction. In particular, it is suitable for group mobility pattern where group partitions and mergence are prevalent behaviors of mobile groups. The proposed clustering scheme aims to form more stable clusters by prolonging cluster lifetime and reducing the clustering iterations even in highly dynamic environment. Simulation results show that the performance of the proposed framework is superior to two widely referenced clustering approaches, the Lowest-ID clustering scheme and the mobility based clustering algorithm MOBIC, in terms of average clusterhead lifetime, average resident time, average number of clusterhead changes, and average number of cluster reaffiliations.

A NONPARAMETRIC MULTI-SEED DATA CLUSTERING TECHNIQUE

ABSTRACT Clustering of data around one seed does not work well if the shape of the cluster is elongated or non-convex. A complex shaped cluster requires several seeds. This study developed a nonparametric multi-seed data clustering approach which splits and merges procedures to handle the complex shapes of clusters. The splitting process utilizes a genetic algorithm to search for the appropriate cluster centers, which split all data into a considered amount of groups. To assign several seeds into one cluster, an innovative clustering process using a minimal spanning tree and statistics concept was proposed to judge whether a pair of clusters should be merged or separated. Experimental results illustrate the difficulties of one-seed-per-cluster, and also the effectiveness of the proposed clustering scheme.

Top-Down Parameter-Free Clustering of High-Dimensional Categorical Data

A parameter-free, fully-automatic approach to clustering high-dimensional categorical data is proposed. The technique is based on a two-phase iterative procedure, which attempts to improve the overall quality of the whole partition. In the first phase, cluster assignments are given, and a new cluster is added to the partition by identifying and splitting a low-quality cluster. In the second phase, the number of clusters is fixed, and an attempt to optimize cluster assignments is done. On the basis of such features, the algorithm attempts to improve the overall quality of the whole partition and finds clusters in the data, whose number is naturally established on the basis of the inherent features of the underlying data set rather than being previously specified. Furthermore, the approach is parametric to the notion of cluster quality: Here, a cluster is defined as a set of tuples exhibiting a sort of homogeneity. We show how a suitable notion of cluster homogeneity can be defined in the context of high-dimensional categorical data, from which an effective instance of the proposed clustering scheme immediately follows. Experiments on both synthetic and real data prove that the devised algorithm scales linearly and achieves nearly optimal results in terms of compactness and separation.

An Entropy-Based Clustering Scheme in Mobile Ad Hoc Networks

The distributiveness of mobile ad hoc networks makes resource allocation strategies very challenging since there is no central node to monitor and coordinate the activities of all the nodes in the network. Since a single node cannot be delegated to act as a centralized authority because of limitations in the transmission range, several delegated nodes may coordinate the activities in certain zones. This methodology is generally referred to as clustering and the nodes are called clusterheads. The clusterheads employ centralized algorithms in its cluster; however, the clusterheads themselves are distributive in nature. In this paper, we propose a clustering scheme i.e., identify a subset of nodes among all the nodes that are best suited to be clusterheads. Though there are several clustering algorithms previously proposed; however, to the best of our knowledge, there is none that characterizes the different node parameters in terms of an information theoretic metric. We use entropy as a measure of local and mutual information available to every node. We considered three parameters in the selection procedure, namely, mobility, energy, and degree. Extensive simulations have been conducted and the performance of the proposed clustering scheme has been compared with the Highest Degree and Lowest ID heuristics in terms of the average number of clusters, the average number of cluster changes, and the average connectivity. The results demonstrate that the mutual information captured through entropy is very effective in determining the most suitable clusterheads.

Size‐restricted cluster formation and cluster maintenance technique for mobile ad hoc networks

AbstractMobile ad hoc networks may be composed of a large number of nodes and hence a hierarchical cluster‐based structure can be employed to address the scalability issues of the large network. In this paper we propose a size‐restricted, distributed clustering strategy (cluster formation and cluster maintenance) for mobile ad hoc networks. A distributed approach where every node is responsible for the clustering decisions would avoid single‐point bottleneck failures. We use a size restriction S, for the cluster formation and cluster maintenance. In addition, while forming the cluster we also use a diameter restriction K. The simulations show that our strategy gives rise to a lesser number of clusters when compared to other clustering algorithms proposed by Gerla et al., Fernadess et al. and Lin et al. This is attributed to the size restriction and cluster merging that we have incorporated. The size restriction helps in better management of resources inside the cluster. Our clustering technique uses a weight‐based cluster head election strategy which results in less change in cluster head and higher cluster head lifetime when compared to the Least Cluster Head Changes (LCC) with lowest ID algorithm. The low change in cluster head and higher cluster head lifetime make our proposed algorithm suitable for high‐mobility situations. The proposed clustering strategy could be used for building the clusters required in a hierarchical structure of the ad hoc network. Copyright © 2007 John Wiley & Sons, Ltd.

Spatial kernel K-harmonic means clustering for multi-spectral image segmentation

The problem of image segmentation using intensity clustering approaches has been addressed in the literature. Grouping pixels of similar intensity to form clusters in an image have been tackled using a number of methods, such as the K-means (KM) algorithm. The K-harmonic means (KHM) was proposed to overcome the sensitivity of KM to centre initialisation. The use of a spatial kernel-based KHM (SKKHM) algorithm on the problem of image segmentation has been investigated. Instead of the original Euclidean intensity distance, a robust kernel-based KHM metric is employed to reduce the effect of outliers and noise. Spatial image information is also incorporated in the proposed clustering scheme, derived from Markov random field modelling. An extension of the proposed algorithm to multi-spectral imaging applications is also presented. Experimental results for both single-channel and multi-channel images demonstrate the robust performance of the proposed SKKHM algorithm.

Two-stage clustering via neural networks

This paper presents a two-stage approach that is effective for performing fast clustering. First, a competitive neural network (CNN) that can harmonize mean squared error and information entropy criteria is employed to exploit the substructure in the input data by identifying the local density centers. A Gravitation neural network (GNN) then takes the locations of these centers as initial weight vectors and undergoes an unsupervised update process to group the centers into clusters. Each node (called gravi-node) in the GNN is associated with a finite attraction radius and would be attracted to a nearby centroid simultaneously during the update process, creating the Gravitation-like behavior without incurring complicated computations. This update process iterates until convergence and the converged centroid corresponds to a cluster. Compared to other clustering methods, the proposed clustering scheme is free of initialization problem and does not need to pre-specify the number of clusters. The two-stage approach is computationally efficient and has great flexibility in implementation. A fully parallel hardware implementation is very possible.

Knowledge-based clustering scheme for collection management and retrieval of library books

We propose a knowledge-based clustering scheme for grouping books in a library. Such a grouping is achieved with the help of domain knowledge in the form of the ACM CR (Computing Reviews) category hierarchy. A new knowledge-based similarity measure is defined and used in clustering books. The proposed scheme is useful in overcoming several problems associated with the existing book collection management and document retrieval systems. More specifically, it can be used in: (1) helping the user select an appropriate collection of books in a library which contains the topics of interest; (2) assigning a classification number to a new book; (3) designing a more appropriate and uniform classification scheme for books; and (4) comparison of libraries based on their collections. Initial experiments on a collection of hundred books using the proposed clustering scheme have given us encouraging results.