Cluster Validation Techniques Research Articles

Leakage detection in water distribution networks using machine-learning strategies

Abstract This work proposes a reliable leakage detection methodology for water distribution networks (WDNs) using machine-learning strategies. Our solution aims at detecting leakage in WDNs using efficient machine-learning strategies. We analyze pressure measurements from pumps in district metered areas (DMAs) in Stockholm, Sweden, where we consider a residential DMA of the water distribution network. Our proposed methodology uses learning strategies from unsupervised learning (K-means and cluster validation techniques), and supervised learning (learning vector quantization algorithms). The learning strategies we propose have low complexity, and the numerical experiments show the potential of using machine-learning strategies in leakage detection for monitored WDNs. Specifically, our experiments show that the proposed learning strategies are able to obtain correct classification rates up to 93.98%.

Recommendations for validating hierarchical clustering in consumer sensory projects

Choosing the proper hierarchical clustering algorithm and number of clusters is always a key question in consumer sensory projects. In many cases, researchers do not publish any reason why it was chosen a given distance measure and linkage rule along with cluster numbers. The reason behind this could be that different cluster validation and comparison techniques give contradictory results in most cases. A complex evaluation to define the proper clustering might be time-consuming and tedious. The paper introduces the clustering of three sensory data sets using different distance metrics and linkage rules for different numbers of clusters. The results of the validation methods deviate, suggesting that clustering depends heavily on the data set in question. Although Euclidean distance, Ward's method seems a safe choice, testing, and validation of different clustering combinations is strongly suggested.

Multicriterion decision making in groundwater planning

Abstract The groundwater planning problems are often multiobjective. Due to conflicting objectives and non-linearity of the variables involved, several feasible solutions may have to be evolved rather than single optimal solution. In this study, the simulation model built on an Analytic Element Method (AEM) and the optimization model built on a Non-dominated Sorting Genetic Algorithm (NSGA-II) were coupled and applied to study a part of the Dore river catchment, France. The maximization of discharge, the minimization of pumping cost and the minimization of piping cost are the three objectives considered. 2105 non-dominated groundwater planning strategies were generated. K-Means cluster analysis was employed to classify the strategies, and clustering was performed for 3 to 25 clusters. A cluster validation technique, namely Davies–Bouldin (DB) index, was employed to find the optimal number of clusters of groundwater strategies which were found to be 20. Multicriterion Decision-Making (MCDM) techniques, namely VIKOR and TOPSIS, were developed to rank the 20 representative strategies. Both these decision-making techniques preferred representative strategy A5 (piping cost, pumping cost and discharge respectively of 880,000 Euro, 679,000 Euro and 1,263.1 m3/s). The sensitivity analysis of parameter v in VIKOR suggested that there were changes in ranking pattern for various values of v. However, the first position remained unchanged.

Key-Frame Extraction Based on HSV Histogram and Adaptive Clustering

Along with the fast development of digital information technology and the application of Internet, video data begins to grow explosively. Some applications with high real-time requirements, such as object detection, require strong online video storage and analysis capabilities. Key-frame extraction is an important technique in video analysis, which provides an organizational framework for dealing with video content and reduces the amount of data required in video indexing. To address the problem, this study proposes a key-frame extraction method based on HSV (hue, saturation, value) histogram and adaptive clustering. The HSV histogram is used as color features for each frame, which reduces the amount of data. Furthermore, by using the transformed one-dimensional eigenvector, the fixed number of features can be extracted for images with different sizes. Then, a cluster validation technique, the silhouette coefficient, is employed to get the appropriate number of clusters without setting any clustering parameters. Finally, several algorithms are compared in the experiments. The density peak clustering algorithm (DPCA) model is shown to be more effective than the other four models in precision and F-measure.

Evaluation of organizational structure through cluster validation analysis of email communications

In this work, we apply cluster validation measures for analyzing email communications at an organizational level of a company. This analysis can be used to evaluate the company structure and to produce further recommendations for structural improvements. Our evaluations, based on data in the forms of email logs and organizational structure for a large European telecommunication company, show that cluster validation techniques can be useful tools for assessing the organizational structure using objective analysis of internal email communications, and for simulating and studying different reorganization scenarios.

Clustering of short time-course gene expression data with dissimilar replicates

Microarrays are used in genetics and medicine to examine large numbers of genes simultaneously through their expression levels under any condition such as a disease of interest. The information from these experiments can be enriched by following the expression levels through time and biological replicates. The purpose of this study is to propose an algorithm which clusters the genes with respect to the similarities between their behaviors through time. The algorithm is also aimed at highlighting the genes which show different behaviors between the replicates and separating the constant genes that keep their baseline expression levels throughout the study. Finally, we aim to feature cluster validation techniques to suggest a sensible number of clusters when it is not known a priori. The illustrations show that the proposed algorithm in this study offers a fast approach to clustering the genes with respect to their behavior similarities, and also separates the constant genes and the genes with dissimilar replicates without any need for pre-processing. Moreover, it is also successful at suggesting the correct number of clusters when that is not known.

Eratosthenes sieve based key-frame extraction technique for event summarization in videos

The rapid growth of video data demands both effective and efficient video summarization methods so that users are empowered to quickly browse and comprehend a large amount of video content. It is a herculean task to manage access to video content in real time where humongous amount of audiovisual recorded data is generated every second. In this paper we propose an Eratosthenes Sieve based key-frame extraction approach for video summarization (VS) which can work better for real-time applications. Here, Eratosthenes Sieve is used to generate sets of all Prime number frames and nonprime number frames up to total N frames of a video. k-means clustering procedure is employed on these sets to extract the key–frames quickly. Here, the challenge is to find the optimal set of clusters, achieved by employing Davies-Bouldin Index (DBI). DBI a cluster validation technique which allows users with free parameter based VS approach to choose the desired number of key-frames without incurring additional computational costs. Moreover, our proposed approach includes likes of both local and global perspective videos. The method strongly enhances clustering procedure performance trough engagement of Eratosthenes Sieve. Qualitative and quantitative evaluation and complexity computation are done in order to compare the performances of the proposed model and state-of-the-art models. Experimental results on two benchmark datasets with various types of videos exhibit that the proposed methods outperform the state-of-the-art models on F-measure.

Cluster analysis of molecular simulation trajectories for systems where both conformation and orientation of the sampled states are important.

Clustering methods have been widely used to group together similar conformational states from molecular simulations of biomolecules in solution. For applications such as the interaction of a protein with a surface, the orientation of the protein relative to the surface is also an important clustering parameter because of its potential effect on adsorbed-state bioactivity. This study presents cluster analysis methods that are specifically designed for systems where both molecular orientation and conformation are important, and the methods are demonstrated using test cases of adsorbed proteins for validation. Additionally, because cluster analysis can be a very subjective process, an objective procedure for identifying both the optimal number of clusters and the best clustering algorithm to be applied to analyze a given dataset is presented. The method is demonstrated for several agglomerative hierarchical clustering algorithms used in conjunction with three cluster validation techniques. © 2016 Wiley Periodicals, Inc.

Time Frequency Single Source Point Detection Based Method for Radar Signal Sorting

Common sorting method have low sorting rates and is sensitive to the Signal-to-Noise Ratio(SNR), detection of time frequency single source point is applied to sort unknown complicated radar signal, high sorting accuracy can be got. The single source points of each radar source signal is detected, then the mixing vector in the corresponding single source point set was estimated by Singular Value Decomposition (SVD), the mixed matrix is estimated simultaneously by cluster validation technique, based on k-means clustering algorithm. Finally, Pulse Description Words (PDW) of each radar signal can be worked out. Experiment results demonstrated that the radar emitter signals extracted by this method showed good performance of noise-resistance and clustering at large-scale SNR. Radar recommaissance is a key part of electromagnetic countermeasure. The mixed signal is sorted by reconnaissance system and the pulse description words of each signal pulse can be calculated to locate radiation sources or assess the threat of each source. In (1), Blind Signal Separation (BSS) was applied in radar signal sorting, a method based on fourth-order cumulant was proposed to accomplish signal sorting and achieved good results, but the number of source signals had been already known in that simulation experiment, the circumstances of unknown signal numbers was not mentioned in that paper. In (2), another blind separation method based on fixed point independent component analysis algorithm was proposed to solve the problem of overdetermined radar signal sorting, but the method can not accomplish underdetermined radar signal sorting. In order to solve the problem of unknown signal numbers in radar signal sorting by BSS, too many methods have been presented. In (3), a robust algorithm under the condition of unknown signal source number was proposed. In that algorithm, the data received by the array was pretreated via the projection transformation to inhibit model errors and reduce data dimension,thus improving the robustness and lightening the computation load. Then, the number of signal could be estimated according to the transformed m-Capon spatial spectrum function. In (4), the author took advantage of the straight line clustering of the sparse source signals, standardized the aliasing signals, then the aliasing signals were formed spherical cluster, thus the linear cluster was turned into density cluster. And the clustering center was searched and obtained by using the ant clustering algorithm, the aliasing matrix can be accurately evaluated. In a word, most methods require clustering to avoid the problem of unknown signal numbers, thus a good clustering algorithm is the key to solve this problem. The problem of underdetermined signal sorting can be settled by two methods: 1. Using the sparsity of source signals; 2. 'Two-step approach'. Both of them work under the condition of sparse signal. In the

A new improved cluster validity indexing technique: harnessed from Goodman-Kruskal validity index

The true potential of clustering techniques is not harnessed optimally because of several reasons. Clustering is implemented either on the preclassified datasets or if implemented on unclassified datasets, it remains unacceptable because its validity cannot be established. Cluster validity techniques come to rescue in the latter cases. Several internal and external cluster validity indices are studied and used to validate the clustering techniques. Moreover, the validity of the indexing techniques is needed to be established first. The current work suggests an improvement over the Goodman-Kruskal indexing technique and establishes its validity by applying it on several benchmark datasets. Hence, it suggests a new cluster validity indexing technique.

Empirical development taxonomy of micro, small and medium-sized enterprises in South Moravian Region

This paper represents a pilot study examining entrepreneurial behavior patterns within the area of development of micro, small and medium-sized market active enterprises in South Moravian Region from the perspective of strategic management. The main aim of this article is to classify the enterprises by the character of development activities into relatively homogeneous groups and identify those activities that contribute to considerable differences between the individual groups. Empirical evidence is based on a quantitative survey in terms of a random choice of respondents obtained by means of an e-mail questionnaire. The major analytic procedure applied within this research comprises methods of Cluster Analysis and validation technique based on chi-square and G-square nonparametric tests for testing differences (independence) of variables between the clusters. By means of Cluster Analysis three relatively homogenous SME groups were identified. The first cluster, called “global”, (approximately 6% of the respondents) prevails in most of the strategic development aspects of the researched enterprises. Global SMEs rely on development activities in the area of product innovations and risk diversification into other activities. In terms of changes in functional strategies this involves primarily the area of new technology and human resources development, and the ability to benefit from participation in networks of cooperating organizations. The second cluster, called “transition”, (approximately 59% of the respondents) is developed mainly by means of investments in business specific knowledge, and/or in technical skills of their employees. They are more conservative in the area of product innovations and risk diversification into other activities, which is then reflected in a lower level of investments in technologies and marketing. In comparison with global enterprises the major difference lies in a much lower number of activities in the area of seeking cooperation with other enterprises. The third cluster, called “local”, (approximately 35% of the respondents) chooses the way of product innovations whose supply is mostly composed of standardized products and services, especially in local or regional markets. Local enterprises, when compared to the former clusters, are typical for a lower level of business cooperation, and in fact for absence of utilizing the support programs.

Reporting and analyzing alternative clustering solutions by employing multi-objective genetic algorithm and conducting experiments on cancer data

Clustering is an essential research problem which has received considerable attention in the research community for decades. It is a challenge because there is no unique solution that fits all problems and satisfies all applications. We target to get the most appropriate clustering solution for a given application domain. In other words, clustering algorithms in general need prior specification of the number of clusters, and this is hard even for domain experts to estimate especially in a dynamic environment where the data changes and/or become available incrementally. In this paper, we described and analyze the effectiveness of a robust clustering algorithm which integrates multi-objective genetic algorithm into a framework capable of producing alternative clustering solutions; it is called Multi-objective K-Means Genetic Algorithm (MOKGA). We investigate its application for clustering a variety of datasets, including microarray gene expression data. The reported results are promising. Though we concentrate on gene expression and mostly cancer data, the proposed approach is general enough and works equally to cluster other datasets as demonstrated by the two datasets Iris and Ruspini. After running MOKGA, a pareto-optimal front is obtained, and gives the optimal number of clusters as a solution set. The achieved clustering results are then analyzed and validated under several cluster validity techniques proposed in the literature. As a result, the optimal clusters are ranked for each validity index. We apply majority voting to decide on the most appropriate set of validity indexes applicable to every tested dataset. The proposed clustering approach is tested by conducting experiments using seven well cited benchmark data sets. The obtained results are compared with those reported in the literature to demonstrate the applicability and effectiveness of the proposed approach.

A Novel Decision Cluster Classifier with Nested Agglomerative K-Means

Classification is an important task in machine learning community. The basic assumption of classification is that the objects that are spatially close in data space tend to have the same label, which is quite similar to the objective of clustering, i.e. grouping similar objects into the same cluster. Based on this fact, several researchers have proposed to build classifiers by using the K-means type algorithms to partition the data sequentially, which are named as decision cluster classifiers. While it is well-known that the performance of K-means type algorithms relies on the locations of initial centers and the selections of number of clusters, it is necessary to deal with both issues well such that the clustering performance is good enough to build such classifiers. In this paper, we propose to build decision cluster classifiers by using the agglomerative fuzzy K-means because of its three good properties: (1) the clustering algorithm is insensitive to initial centers; (2) the algorithm can automatically determine the number of clusters combined with cluster validation techniques; (3) the algorithm can control the density level of clusters identified. The experimental results on UCI benchmark data sets have shown the effectiveness of our proposed approach.

Pattern recognition approach to identify natural clusters of acoustic emission signals

A new approach is introduced to identify natural clusters of acoustic emission signals. The presented technique is based on an exhaustive screening taking into account all combinations of signal features extracted from the recorded acoustic emission signals. For each possible combination of signal features an investigation of the classification performance of the k-means algorithm is evaluated ranging from two to ten classes. The numerical degree of cluster separation of each partition is calculated utilizing the Davies–Bouldin and Tou indices, Rousseeuw’s silhouette validation method and Hubert’s Gamma statistics. The individual rating of each cluster validation technique is cumulated based on a voting scheme and is evaluated for the number of clusters with best performance. This is defined as the best partitioning for the given signal feature combination. As a second step the numerical ranking of all these partitions is evaluated for the globally optimal partition in a second voting scheme using the cluster validation methods results. This methodology can be used as an automated evaluation of the number of natural clusters and their partitions without previous knowledge about the cluster structure of acoustic emission signals. The suitability of the current approach was evaluated using artificial datasets with defined degree of separation. In addition the application of the approach to clustering of acoustic emission signals is demonstrated for signals obtained from failure during loading of carbon fiber reinforced plastic specimens.

LUNG SEGMENTATION FOR CHEST RADIOGRAPH BY USING ADAPTIVE ACTIVE SHAPE MODELS

In this article, we propose an automatic lung segmentation method. We design an ROI-adjusted initial boundary derivation method to estimate a proper initial lung boundary for ASM deformation by deriving the translation and the scaling parameters from the lung ROI. An adaptive ASM, using k-means clustering and silhouette-based cluster validation technique, is proposed to adapt to the lung shape change so that the lung shape variation among people can be overwhelmed. The experimental results indicate that the performance of the proposed lung segmentation method is superior to the traditional ASM approach.

ETE: a python Environment for Tree Exploration

BackgroundMany bioinformatics analyses, ranging from gene clustering to phylogenetics, produce hierarchical trees as their main result. These are used to represent the relationships among different biological entities, thus facilitating their analysis and interpretation. A number of standalone programs are available that focus on tree visualization or that perform specific analyses on them. However, such applications are rarely suitable for large-scale surveys, in which a higher level of automation is required. Currently, many genome-wide analyses rely on tree-like data representation and hence there is a growing need for scalable tools to handle tree structures at large scale.ResultsHere we present the Environment for Tree Exploration (ETE), a python programming toolkit that assists in the automated manipulation, analysis and visualization of hierarchical trees. ETE libraries provide a broad set of tree handling options as well as specific methods to analyze phylogenetic and clustering trees. Among other features, ETE allows for the independent analysis of tree partitions, has support for the extended newick format, provides an integrated node annotation system and permits to link trees to external data such as multiple sequence alignments or numerical arrays. In addition, ETE implements a number of built-in analytical tools, including phylogeny-based orthology prediction and cluster validation techniques. Finally, ETE's programmable tree drawing engine can be used to automate the graphical rendering of trees with customized node-specific visualizations.ConclusionsETE provides a complete set of methods to manipulate tree data structures that extends current functionality in other bioinformatic toolkits of a more general purpose. ETE is free software and can be downloaded from http://ete.cgenomics.org.

A data mining approach to dinoflagellate clustering according to sterol composition: correlations with evolutionary history

This study examined the sterol compositions of 102 dinoflagellates using clustering and cluster validation techniques, as a means of determining the relatedness of the organisms. In addition, dinoflagellate sterol-based relationships were compared statistically to 18S rDNA-based phylogenetic relationships using the Mantel test. Our results indicated that the examined dinoflagellates formed six clusters based on sterol composition and that several, but not all, dinoflagellate genera, which formed discrete clusters in the 18S rDNA-based phylogeny, shared similar sterol compositions. This and other correspondences suggest that the sterol compositions of dinoflagellates are explained, to a certain extent, by the evolutionary history of this lineage.

Fuzzy clustering for seafloor classification

In order to develop quantitative seafloor sediment classification techniques it is important to acknowledge that by nature the boundaries between soft sediments are characterized by transition zones and therefore are indeterminate and gradual. A fuzzy clustering method, fuzzy c-means (FCM), was used to identify these transition zones within a subset of the data used to generate the Australian Seascapes classification model. The overlapping classes and gradual boundaries resulting from the fuzzy c-means algorithm provided estimates of sediment boundaries that are a closer model of reality than sharp boundaries. FCM output is given in the form of membership layers for each class, hard classes for each grid cell based on the maximum membership value, and a confusion index layer quantifying uncertainty in class attribution. The confusion index layer provided a spatial representation of transition zones and overlap between seafloor classes and highlighted areas of greatest uncertainty. We extended the standard FCM algorithm by applying the new FMLE fuzzy clustering algorithm that takes into account spatial relationships in the data. In addition, we implemented and applied new cluster validity techniques, PCAES, PBMF, and XB to determine the optimal number of clusters in the data, which is a novel pattern recognition application for seabed mapping. The 5-class FCM classification provided the most reliable result. The results of this research were tested and validated on a simulated dataset and then the clustering and validation algorithms were applied to marine sediment data to identify Seascapes. The new results were compared with previously published Seascapes classes identified with hard ISODATA clustering techniques from GeoScience Australia's Seascapes classification result. With the increasing use of physical surrogates to explain marine biodiversity, this research plays a crucial role in the development of techniques to identify habitat zones on the seabed.

Agglomerative Fuzzy K-Means Clustering Algorithm with Selection of Number of Clusters

In this paper, we present an agglomerative fuzzy K-means clustering algorithm for numerical data, an extension to the standard fuzzy K-means algorithm by introducing a penalty term to the objective function to make the clustering process not sensitive to the initial cluster centers. The new algorithm can produce more consistent clustering results from different sets of initial clusters centers. Combined with cluster validation techniques, the new algorithm can determine the number of clusters in a data set, which is a well known problem in $k$-means clustering. Experimental results on synthetic data sets (2 to 5 dimensions, 500 to 5000 objects and 3 to 7 clusters), the BIRCH two-dimensional data set of 20000 objects and 100 clusters, and the WINE data set of 178 objects, 17 dimensions and 3 clusters from UCI, have demonstrated the effectiveness of the new algorithm in producing consistent clustering results and determining the correct number of clusters in different data sets, some with overlapping inherent clusters.

“Best K”: critical clustering structures in categorical datasets

The demand on cluster analysis for categorical data continues to grow over the last decade. A well-known problem in categorical clustering is to determine the best K number of clusters. Although several categorical clustering algorithms have been developed, surprisingly, none has satisfactorily addressed the problem of best K for categorical clustering. Since categorical data does not have an inherent distance function as the similarity measure, traditional cluster validation techniques based on geometric shapes and density distributions are not appropriate for categorical data. In this paper, we study the entropy property between the clustering results of categorical data with different K number of clusters, and propose the BKPlot method to address the three important cluster validation problems: (1) How can we determine whether there is significant clustering structure in a categorical dataset? (2) If there is significant clustering structure, what is the set of candidate “best Ks”? (3) If the dataset is large, how can we efficiently and reliably determine the best Ks?