Search

The purpose of this report is to investigate current existing algorithm to cluster sequential data based on hidden Markov model (HMM). Clustering is a classic technique that divides a set of objects into groups (called clusters) so that objects in the same cluster are similar in some sense. The clustering of sequential or time series data, however, draws lately more and more attention from researchers. Hidden Markov model (HMM)-based clustering of sequences is probabilistic model-based approach to clustering sequences. Generally, there are two kinds of methodologies: parametric and semi-parametric. The parametric methods make strict assumptions that each cluster is represented by a corresponding HMM, while the semi-parametric approaches relax this assumption and transform the problem to a similarity-based issue. Generally, the semi-parametric methods perform better than parametric approaches as reported by some researchers. Future research can be done in exploring new distance measures between sequences and extending current HMM-based methodologies by using other models.

In many popular applications like peer-to-peer systems, large amounts of data are distributed among multiple sources. Analysis of this data and identifying clusters is challenging due to processing, storage, and transmission costs. In this paper, we propose GDCluster, a general fully decentralized clustering method, which is capable of clustering dynamic and distributed data sets. Nodes continuously cooperate through decentralized gossip-based communication to maintain summarized views of the data set. We customize GDCluster for execution of the partition-based and density-based clustering methods on the summarized views, and also offer enhancements to the basic algorithm. Coping with dynamic data is made possible by gradually adapting the clustering model. Our experimental evaluations show that GDCluster can discover the clusters efficiently with scalable transmission cost, and also expose its supremacy in comparison to the popular method LSP2P.

In Hierarchical clustering, a set of patterns are partitioned into a sequence of groups represented as a dendrogram. The dendrogram is a tree representation where each node is associated with merging of two (or more) partitions and hence each partition is nested into the next partition. Hierarchical representation has properties that are useful for visualization and interpretation of clustering results. On one hand, different hierarchical clustering algorithms usually produce different dendrograms. On the other hand, clustering combination methods have received considerable interest in recent years and they yield superior results for clustering problems. This paper proposes a framework for combining various hierarchical clustering results which preserves the structural contents of input hierarchies. In this method, first a description matrix is created for each hierarchy, and then the description matrices of the input hierarchies are aggregated to form a consensus matrix from which the final hierarchy is derived. In this framework, we use two new measures for aggregating the description matrices, namely Renyi and Jensen–Shannon Divergences. The experimental and comparative analysis of our proposed framework shows the effectiveness of these two aggregators in hierarchical clustering combination.

Most of the researches on text categorization are focus on using bag of words. Some researches provided other methods for classification such as term phrase, Latent Semantic Indexing, and term clustering. Term clustering is an effective way for classification, and had been proved as a good method for decreasing the dimensions in term vectors. The authors used hierarchical term clustering and aggregating similar terms. In order to enhance the performance, they present a modify indexing with terms in cluster. Their test collection extracted from Chinese NETNEWS, and used the Centroid-Based classifier to deal with the problems of categorization. The results had shown that term clustering is not only reducing the dimensions but also outperform than bag of words. Thus, term clustering can be applied to text classification by using any large corpus, its objective is to save times and increase the efficiency and effectiveness. In addition to performance, these clusters can be considered as conceptual knowledge base, and kept related terms of real world.

Social tagging is one of the vital attributes of WEB2.0. The challenge of Web 2.0 is a gigantic measure of information created over a brief time. Tags are broadly used to interpret and arrange the web 2.0 assets. Tag clustering is the procedure of grouping the comparable tags into clusters. The tag clustering is extremely valuable for researching and organizing the web2. 0 resources furthermore critical for the achievement of Social Bookmarking frameworks. In this paper, the authors proposed a hybrid Tolerance Rough Set Based Firefly TRS-Firefly-K-Means clustering algorithm for clustering tags in social systems. At that stage, the proposed system is contrasted with the benchmark algorithm K-Means clustering and Particle Swarm optimization PSO based Clustering technique. The experimental analysis outlines the viability of the suggested methodology.

© 2017 Elsevier B.V. This paper presents a comprehensive study on clustering: exiting methods and developments made at various times. Clustering is defined as an unsupervised learning where the objects are grouped on the basis of some similarity inherent among them. There are different methods for clustering the objects such as hierarchical, partitional, grid, density based and model based. The approaches used in these methods are discussed with their respective states of art and applicability. The measures of similarity as well as the evaluation criteria, which are the central components of clustering, are also presented in the paper. The applications of clustering in some fields like image segmentation, object and character recognition and data mining are highlighted.

Rare efforts have been devoted to generating the structured Navigation Box (Navbox) for Wikipedia articles. A Navbox is a table in Wikipedia article page that provides a consistent navigation system for related entities. Navbox is critical for the readership and editing efficiency of Wikipedia. In this paper, we target on the automatic generation of Navbox for Wikipedia articles. Instead of performing information extraction over unstructured natural language text directly, an alternative avenue is explored by focusing on a rich set of semi-structured data in Wikipedia articles: linked entities. The core idea of this paper is as follows: If we cluster the linked entities and interpret them appropriately, we can construct a high-quality Navbox for the article entity. We propose a clustering-then-labeling algorithm to realize the idea. Experiments show that the proposed solutions are effective. Ultimately, our approach enriches Wikipedia with 1.95 million new Navboxes of high quality.

In recent years, various data clustering algorithms have been proposed in the data mining and engineering communities. However, there are still drawbacks in traditional clustering methods which are worth to be further investigated, such as clustering for the high dimensional data, learning an ideal affinity matrix which optimally reveals the global data structure, discovering the intrinsic geometrical and discriminative properties of the data space, and reducing the noises influence brings by the complex data input. In this paper, we propose a novel clustering algorithm called robust dual clustering with adaptive manifold regularization (RDC), which simultaneously performs dual matrix factorization tasks with the target of an identical cluster indicator in both of the original and projected feature spaces, respectively. Among which, the $l_{2,1}$ -norm is used instead of the conventional $l_{2}$ -norm to measure the loss, which helps to improve the model robustness by relieving the influences by the noises and outliers. In order to better consider the intrinsic geometrical and discriminative data structure, we incorporate the manifold regularization term on the cluster indicator by using a particularly learned affinity matrix which is more suitable for the clustering task. Moreover, a novel augmented lagrangian method (ALM) based procedure is designed to effectively and efficiently seek the optimal solution of the proposed RDC optimization. Numerous experiments on the representative data sets demonstrate the superior performance of the proposed method compares to the existing clustering algorithms.

In previous works, we proposed two methods for data clustering based on automatically discovered fuzzy predicates which were referred to as SOM-based Fuzzy Predicate Clustering (SFPC) [Meschino et al., Neurocomputing, 147, 47–59 (2015)] and Type-2 Data-based Fuzzy Predicate Clustering (T2-DFPC) [Comas et al., Expert Syst. Appl., 68, 136–150 (2017)]. In such methods, fuzzy predicates allow both data clustering and knowledge discovering about the obtained clusters. This last feature constitutes novelty comparing to other existing approaches and it is a major contribution in the data clustering field. Based on these previous methods, in the present paper a new automatic clustering method based on fuzzy predicates is proposed which uses Self-Organizing Maps (SOMs) and is called Type-2 SOM-based Fuzzy Predicate Clustering (T2-SFPC). The new method does not require any prior knowledge about the clustering addressed. First, a random partition is defined on the dataset to be clustered and SOMs are configured and trained using the resulting data subsets. Second, an automatic clustering approach is applied on the SOM codebooks, discovering representative data of the different clusters, which are called cluster prototypes. Third, interval type-2 membership function formed by Gaussian-shape sub-functions and fuzzy predicates are defined, allowing data clustering and its interpretation. The proposed method preserves all the advantages of the previous methods SFPC and T2-DFPC in relation to the knowledge extraction capabilities and their potential application on distributed clustering and parallel computing, but results obtained on several public datasets tested showed more compactness and separation of the clusters defined by the T2-SFPC, outperforming both the previous methods and the several classical clustering approaches tested, considering internal and external validation indices. Additionally, both clustering interpretation and optimization capabilities are improved by the proposed method when compared to the methods SFPC and T2-DFPC. Fil: Comas, Diego Sebastián. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Mar del Plata. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica. Universidad Nacional de Mar del Plata. Facultad de Ingeniería. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica.; Argentina Fil: Pastore, Juan Ignacio. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Mar del Plata. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica. Universidad Nacional de Mar del Plata. Facultad de Ingeniería. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica.; Argentina Fil: Bouchet, Agustina. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Mar del Plata. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica. Universidad Nacional de Mar del Plata. Facultad de Ingeniería. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica.; Argentina Fil: Ballarin, Virginia Laura. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Mar del Plata. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica. Universidad Nacional de Mar del Plata. Facultad de Ingeniería. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica.; Argentina Fil: Meschino, Gustavo Javier. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Mar del Plata. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica. Universidad Nacional de Mar del Plata. Facultad de Ingeniería. Instituto de Investigaciones Científicas y Tecnológicas en Electrónica.; Argentina. Universidad Nacional de Mar del Plata. Facultad de Ingeniería. Departamento de Ingeniería Eléctrica. Laboratorio de Bioingeniería; Argentina

In this paper a novel approach to fuzzy hierarchical clustering of short text fragments is presented. Nowadays dataset which contains a large and even huge amount of short text fragments becomes quite a common object. Different kinds of short messages, paper or news headers are examples of this kind of objects. Authors have taken another similar object which is a dataset of key process indicators of Strategic Planning System of Russian Federation.

© 2017 IEEE. Document clustering groups documents of certain similar characteristics in one cluster. Document clustering has shown advantages on organization, retrieval, navigation and summarization of a huge amount of text documents on Internet. This paper presents a novel, unsupervised approach for clustering single-author documents into groups based on authorship. The key novelty is that we propose to extract contextual correlations to depict the writing style hidden among sentences of each document for clustering the documents. For this purpose, we build an Hidden Markov Model (HMM) for representing the relations of sequential sentences, and a two-level, unsupervised framework is constructed. Our proposed approach is evaluated on four benchmark datasets, widely used for document authorship analysis. A scientific paper is also used to demonstrate the performance of the approach on clustering short segments of a text into authorial components. Experimental results show that the proposed approach outperforms the state-of-the-art approaches.

In this paper, we present a segmentation system for German texts. We apply conditional random fields (CRF), a statistical sequential model, to a type of text used in private communication. We show that by segmenting individual punctuation, and by taking into account freestanding lines and that using unsupervised word representation (i. e., Brown clustering, Word2Vec and Fasttext) achieved a label accuracy of 96% in a corpus of postcards used in private communication.

The challenge of detecting research topics in a specific research field has attracted attention from researchers in the bibliometrics community. In this study, to solve two problems of clustering papers, i.e., the influence of different distributions of citation links and involved textual features on similarity computation, the authors propose a hybrid self-optimized clustering model to detect research topics by extending the hybrid clustering model to identify "core documents". First, the Amsler network, consisting of bibliographic coupling and co-citation links, is created to calculate the citation-based similarity based on the cosine angle of papers. Second, the cosine similarity is also used to compute the text-based similarity, which consists of the textual statistical and topological features. Then, the cosine angle of the linear combination of citation- and text-based similarity is considered as the hybrid similarity. Finally, the Louvain method is applied to cluster papers, and the terms based on term frequency are used to label clusters. To test the performance of the proposed model, a dataset related to the data envelopment analysis field is used for comparison and analysis of clustering results. Based on the benchmark built, different clustering methods with different citation links or textual features are compared according to evaluation measures. The results show that the proposed model can obtain reasonable and effective clustering results, and the research topics of data envelopment analysis field are also analyzed based on the proposed model. As different features are considered in the proposed model compared with previous hybrid clustering models, the proposed clustering model can provide inspiration for further studies on topic identification by other researchers.

Spectral clustering is one of the most prominent clustering approaches. However, it is highly sensitive to noisy input data. In this work, we propose a robust spectral clustering technique able to handle such scenarios. To achieve this goal, we propose a sparse and latent decomposition of the similarity graph used in spectral clustering. In our model, we jointly learn the spectral embedding as well as the corrupted data - thus, enhancing the clustering performance overall. We propose algorithmic solutions to all three established variants of spectral clustering, each showing linear complexity in the number of edges. Our experimental analysis confirms the significant potential of our approach for robust spectral clustering. Supplementary material is available at www.kdd.in.tum.de/RSC.

The analysis of flow traces can help to understand a network's usage patterns. We present a hierarchical clustering algorithm for network flow data that can summarize terabytes of IP traffic into a parsimonious tree model. The method automatically finds an appropriate scale of aggregation so that each cluster represents a local maximum of the traffic density from a block of source addresses to a block of destination addresses. We apply this clustering method on NetFlow data from an enterprise network, find the largest traffic clusters, and analyze their stationarity across time. The existence of heavy-volume clusters that persist over long time scales can help network operators to perform usage-based accounting, capacity provisioning and traffic engineering. Also, changes in the layout of hierarchical clusters can facilitate the detection of anomalies and significant changes in the network workload.

K-means plays an important role in different fields of data mining. However, k-means often becomes sensitive due to its random seeds selecting. Motivated by this, this article proposes an optimized k-means clustering method, named k*-means, along with three optimization principles. First, we propose a hierarchical optimization principle initialized by k* seeds ([Formula: see text]) to reduce the risk of random seeds selecting, and then use the proposed “top- n nearest clusters merging” to merge the nearest clusters in each round until the number of clusters reaches at [Formula: see text]. Second, we propose an “optimized update principle” that leverages moved points updating incrementally instead of recalculating mean and [Formula: see text] of cluster in k-means iteration to minimize computation cost. Third, we propose a strategy named “cluster pruning strategy” to improve efficiency of k-means. This strategy omits the farther clusters to shrink the adjustable space in each iteration. Experiments performed on real UCI and synthetic datasets verify the efficiency and effectiveness of our proposed algorithm.

Clustering of multimodal data according to their information content is considered in this paper. Statistical correlations present in data that contain similar information are exploited to perform the clustering task. Specifically, multiset canonical correlation analysis is equipped with norm-one regularization mechanisms to identify clusters within different types of data that share the same information content. A pertinent minimization formulation is put forth, while block coordinate descent is employed to derive a batch clustering algorithm which achieves better clustering performance than existing alternatives. Distributed implementations are also considered to cluster spatially clustered data utilizing the alternating direction method of multipliers. Relying on subgradient descent, an online clustering approach is derived which substantially lowers computational complexity compared to the batch approaches. Numerical tests demonstrate that the proposed schemes outperform existing alternatives.

Text clustering, as a significant field in natural language processing, is a key technology of processing and organizing massive text data. In the era of big data, however, the massiveness of data brings great challenge in aspects of time and accuracy of text clustering. This paper focus on the issue of speed and preciseness in text clustering combined with genetic algorithm, feedback and distributed computing. A distributed text clustering method is proposed, and it is based on frequent Itemset. The examination result shows it can find out the global optimal centers more efficiently and make the clustering most accurate.

Sharing information between multiple tasks can enhance the accuracy of human action recognition systems. However, using shared information to improve multi-task human action clustering has never been considered before, and cannot be achieved using existing clustering methods. In this work, we present a novel and effective Multi-Task Information Bottleneck (MTIB) clustering method, which is capable of exploring the shared information between multiple action clustering tasks to improve the performance of individual task. Our motivation is that, different action collections always share many similar action patterns, and exploiting the shared information can lead to improved performance. Specifically, MTIB generally formulates this problem as an information loss minimization function. In this function, the shared information can be quantified by the distributional correlation of clusters in different tasks, which is based on a high-level common vocabulary constructed through a novel agglomerative information maximization method. Extensive experiments on two kinds of challenging data sets, including realistic action data sets (HMDB & UCF50, Olympic & YouTube), and cross-view data sets (IXMAS, WVU), show that the proposed approach compares favorably to the state-of-the-art methods.

Text clustering is one of the text mining tasks that is employed in search engines. Discovering the optimal number of clusters for a dataset or repository is a challenging problem. Various clustering algorithms have been reported in the literature but most of them rely on a pre-defined value of the k clusters. In this study, a variant of Firefly algorithm, termed as FireflyClust, is proposed to automatically cluster text documents in a hierarchical manner. The proposed clustering method operates based on five phases: data pre-processing, clustering, item re-location, cluster selection and cluster refinement. Experiments are undertaken based on different selections of threshold value. Results on the TREC collection named TR11, TR12, TR23 and TR45, showed that the FireflyClust is a better approach than the Bisect K-means, hybrid Bisect K-means and Practical General Stochastic Clustering Method. Such a result would enlighten the directions in developing a better information retrieval engine for this dynamic and fast growing big data era.

Link clustering groups different edges in a graph according to their similarities. Link clustering can reveal the overlapping and hierarchical organizations in a wide spectrum of networks. This work studies how to improve efficiency of link clustering along three dimensions, algorithm, modeling, and parallelization, on multi-core machines. We evaluate the efficiency improved due to each of the three dimensions using word association graphs extracted from a twitter dataset.

Trajectory clustering techniques help discover interesting insights from moving object data, including common routes for people and vehicles, anomalous sub-trajectories, etc. Existing trajectory clustering techniques fail to take in to account the uncertainty present in location data. In this paper, we investigate the problem of clustering trajectory data and propose a novel algorithm for clustering similar full and sub-trajectories together while modeling uncertainty in this data. We describe the necessary pre-processing techniques for clustering trajectory data, namely techniques to discretize raw location data using Possible World semantics to capture the inherent uncertainty in location data, and to segment full trajectories in to meaningful sub-trajectories. As a baseline, we extend the well known K-means algorithm to cluster trajectory data. We then describe and evaluate a new trajectory clustering algorithm, SOM-TC (Self-Organizing Map Based Trajectory Clustering), that is inspired from the self-organizing map technique and is at least 4x faster than the baseline K-means and current density based clustering approaches.

The problem of estimating appropriate number of clusters has been a main and difficult issue in clustering researches. There are different methods for this in hierarchical clustering; a typical approach is to try clustering for different number of clusters, and compare them using a measure to estimate cluster numbers. On the other hand, there is no such method to estimate automatically the number of clusters in agglomerative hierarchical clustering (AHC), since AHC produces a family of clusters with different cluster numbers at the same time using the form of dendrograms. An exception is the Newman method in network clustering, but this method does not have a useful dendrogram output. The aim of the present paper is to propose new methods to automatically estimate the number of clusters in AHC. We show two approaches for this purpose, one is to use a variation of cluster validity measure, and another is to use statistical model selection method like BIC.

The article presents the analysis of the clustering problem formalization and considers possibilities to use the classical methods and bio-inspired methods for solving problems of the cluster analysis. In this paper we do not present full review of the new clustering methods, but identify some trends in the development of cluster analysis and special attention is given to area of bio-inspired methods for clustering. Oscillatory networks can be considered as way that is proposed by nature for solving problems of data mining and especially for cluster analysis. Opportunities are examined for applying the oscillatory networks for general problems of cluster analysis. Have shown general features compared with various traditional methods for cluster analysis.

In the bases of increasing the volume of text information, the dealing with text information has become incredibly complicated. The text clustering is a suitable technique used in dealing with a tremendous amount of text documents by classifying these set of text documents into clusters. Ultimately, text documents hold sparse, non-uniform distribution and uninformative features are difficult to cluster. The text feature selection is a primary unsupervised learning method that is utilized to choose a new subset of informational text features. In this paper, a new algorithm is proposed based on β-hill climbing technique for text feature selection problem to improve the text clustering (B-FSTC). The results of the proposed method for β-hill climbing and original Hill climbing (i.e., H-FSTC) are examined using the k-mean text clustering and compared with each other. Experiments were conducted on four standard text datasets with varying characteristics. Interestingly, the proposed β-hill climbing algorithm obtains superior results in comparison with the other well-regard techniques by producing a new subset of informational text features. Lastly, the β-hill climbing-based feature selection method supports the k-mean clustering algorithm to achieve more precise clusters.

Clustering accuracy of fuzzy clustering is sensitive to the structure of dataset to be studied. Semi-supervised clustering algorithms aim to increase the accuracy under the supervisions of a limited amount of labeled data, but the classification rate is highly dependent on the size of available labeled data. To overcome this disadvantage, we propose a novel semi-supervised clustering based on label propagation. Under our label propagation mechanism, an unlabeled datum propagates an estimated label from two aspects: (1) from its adjacent labeled data; (2) from a previous clustering result. The effects of these estimated labels are controlled by weights indicating their confidence levels. The effectiveness of the proposed model with label propagation scheme are evaluated by several real-life data sets. Experimental results show that accuracy level would increase by applying this learning scheme, compared to other semi-supervised models.

The aim of collaborative clustering is to reveal the common underlying structure of data spread across multiple sites by applying different clustering algorithms and therefore improve the final clustering result. The purpose of this article is to introduce a new collaborative topological clustering approach based on Self-Organizing Maps (SOM) with more flexible structure. The main drawback of the previously proposed collaborative methods is they require a strong condition to carry out the collaboration, i.e. all the map related to each site must be learned with the same number of neurones. An alternative approach to overcome the problem of this dimensionality has been proposed in the case of k-means. The idea consists in modifying databases by adding virtual points which convey clustering information, to change the position of centers of the clustering solution. This approach seems promising because it is very flexible. In this paper we propose a method called VP2SOM (Virtual Points to SOM) which uses virtual points to realize the collaboration between SOM maps with different sizes. We have tested the proposed approach on several datasets and the early results seem promising.

Currently, the data that dealt by traditional text clustering methods is small. Text representation model that text clustering used is traditional vector space model (VSM). The traditional text clustering has defect of low efficiency when processing big data. The quality is bad when using traditional VSM model for text representation. To solve these two problems, this paper puts forward a VSM model for text representation, which is improved by Term Frequency(TF), and using Mahout to text clustering. Experimental results show that use Mahout for text clustering has a higher efficiency than traditional text clustering; this novel VSM model can improve the text clustering quality.

International audience; Multi-source clustering is common data mining taskthe aim of which is to use several clustering algorithms to analyzedifferent aspects of the same data. Well known applications ofmulti-source clustering include horizontal collaborative clusteringand multi-view clustering, where several algorithms combinetheir strengths by exchanging information about their finding onlocal structures with a goal of mutual improvement. However,many of these proposed algorithms and statistical models lackthe capability to detect weak collaborations that may provedetrimental to the global clustering process.In this article, we propose a weighing optimization methodthat will help detecting which algorithms should exchange theirinformation based on the diversity between the different algorithms’solutions.

Multi-View Clustering models can be viewed as a way to extract information from different data representations to improve the clustering accuracy. In multi-view clustering, some views are irrelevant and among the relevant ones, some may be more or less relevant than others. This is why the most part of existing algorithms assign a weight to each view aiming to compute its relevance in the clustering process. However very few algorithms computes also the relevance weight of variables inside each view aiming to achieve automated feature selection. This paper proposes a muti-view hard c-means clustering algorithm with automated computation of weights for both views and variables in such a way that the relevant views as well as the relevant variables in each view are selected for clustering. Compared to previous similar works, an advantage of the proposed method is that, apart the need to know previously the number of clusters, there are no additional parameters to tune. Experiments with benchmark data sets corroborate the usefulness of the proposed method.

This paper presents the results of language clustering in the i-vectors space, a method to determine in an unsupervised manner how many languages are in a data set and which recordings contain the same language. The most dense i-vectors clusters are found using the DBSCAN algorithm in a low dimensional space obtained by the t-SNE method. Quality of clustering for spherical k-means and the proposed method are tested with the data from NIST 2015 i-Vector Challenge. Usefulness of obtained clustering is tested in the challenge evaluation system. The results demonstrate that the proposed method allows to find 109 dense clusters with low impurity for 50 target languages.

The aim of clustering is to group objects which have common features within the group but have dissimilarities with other groups. Clustering algorithms involve finding a common structure without using any label, similarly like other unsupervised methods. Recent studies show that the decision of the ensemble clusters gives more accurate results than any single clustering solution. Besides that, the accuracy and diversity of the ensemble are one of the important factors which effect the overall success of the algorithm. There is a trade off between accuracy and diversity, in other words, you sacrifice one while you increase the performance of the other. On the other hand, the optimum number of clustering solutions is one of the parameters that effect the final result. Recently, finding the best subset of the ensemble clustering solutions by eliminating the redundant solutions has become one of the most challenging problems in the literature. The proposed study here aims to find a best model which optimizes the accuracy and diversity trade off by selecting the best subset of cluster ensemble.

Correlation clustering is a technique for aggregating data based on qualitative information about which pairs of objects are labeled `similar' or `dissimilar.' Because the optimization problem is NP-hard, much of the previous literature focuses on finding approximation algorithms. In this paper we explore how to solve the correlation clustering objective exactly when the data to be clustered can be represented by a low-rank matrix. We prove in particular that correlation clustering can be solved in polynomial time when the underlying matrix is positive semidefinite with small constant rank, but that the task remains NP-hard in the presence of even one negative eigenvalue. Based on our theoretical results, we develop an algorithm for efficiently ``solving'' low-rank positive semidefinite correlation clustering by employing a procedure for zonotope vertex enumeration. We demonstrate the effectiveness and speed of our algorithm by using it to solve several clustering problems on both synthetic and real-world data.

This paper explores the clustering based multi-label classification problem. We propose an overlapping clustering based multi-label classification method (OCBMLC). In the experiment we compare three multi-label classification algorithms that are based on clustering. In the experimental setting section, we use three clustering methods for datasets before multi-label classification. Two popular multi-label benchmark datasets are used in our experiments: emotions dataset and yeast dataset. We employ micro F1-measure and hamming loss as multi-label classification performance evaluation metrics. The experimental results demonstrate that various clustering algorithms generate different performance of the multi-label classification, and the overlapping clustering based multi-label classification algorithm may adapt to multi-label classification more.

We present a novel approach for large speech databases quantization. It uses an unsupervised iterative process to regulate a similarity measure to set the number of clusters and their boundaries, thus overcoming the shortcomings of conventional clustering algorithms such as k-Means and Fuzzy C-Means, which require a priori knowledge of the number of clusters and a similarity measure that follows the data distribution, and which are sensitive to the choice of the initial configuration. The integration of the proposed clustering algorithm into an automatic hybrid HMM/MLP speech recognition system showed improved recognition accuracy in comparison with that obtained by FCM clustering whose outcome was optimized by a genetic algorithm.

The purpose of text mining is to extract meaningful information from documents for various applications. One such application is document clustering. Document clustering refers to the clustering of similar documents into a segment. However, the numerous sentiments, emotions, and concepts involved in documents complicate the task of document clustering. In this paper, we combine the results of topic analysis and sentiment analysis to perform the co-clustering of documents. In contrast to previous papers, the key characteristic of the proposed method is soft clustering, and it considers topics and subjectivity (including emotions and sentiments) simultaneously. The empirical results indicate that the proposed method can effectively cluster documents based on the MANOVA test. In addition, the proposed method also provides a flexible way to co-cluster documents based on topics and/or subjectivity.

Fifty years have gone by since the publication of the first paper on clustering based on fuzzy sets theory. In 1965, L.A. Zadeh had published Fuzzy Sets [335]. After only one year, the first effects of this seminal paper began to emerge, with the pioneering paper on clustering by Bellman, Kalaba, Zadeh [33], in which they proposed a prototypal of clustering algorithm based on the fuzzy sets theory. Starting from this paper, several uncertain clustering methods based on different theoretical approaches for modeling the uncertainty have been proposed. The present paper presents a systematic literature review of these clustering approaches. In particular, with respect to the Statistical Reasoning System, we first illustrate the connection between Information and Uncertainty from the perspective of the so-called Informational Paradigm, according to which Information is constituted by Informational ingredients, specifically the Empirical Information, represented by statistical data, and Theoretical information consisting of background knowledge and basic modeling assumptions. We then describe different kinds of uncertainty affecting the Information. Focusing on the uncertainty associated with a particular statistical methodology, i.e. Cluster Analysis, and adopting as theoretical platform the Informational Paradigm, we present a systematic literature review of different uncertainty-based clustering approaches -i.e. Fuzzy clustering, Possibilistic clustering, Shadowed clustering, Rough sets-based clustering, Intuitionistic fuzzy clustering, Evidential clustering, Credibilistic clustering, Type-2 fuzzy clustering, Neutrosophic clustering, Hesitant fuzzy clustering, Interval-based fuzzy clustering, and Picture fuzzy clustering. We thus show how all these clustering approaches are able of managing in different ways the uncertainty associated with the two components of the Informational Paradigm, i.e. the Empirical and Theoretical Information.

This thesis tries to analyze the drawbacks and shortages of the present ant-based text clustering algorithms (Z-ACTCA algorithm in short). The author attempts to improve the ant-based text clustering algorithms from the following three aspects: text similarity calculation, iterations termination condition as well as parametric adaption. Meanwhile, preprocession will be done on the two-dimensional grid of data mapping and finally the simulation experiment is conducted based on the dataset in the corpus of Chinese text classification in Fu Dan University. At the end of the thesis, the author draws the conclusion that the improved algorithms are superior to the ant-based text clustering algorithms in terms of both the text clustering accuracy and the convergence of text clustering.

These We mean by the clustering is a technique to divide the text into clusters of words, so that words in the same cluster are similar to each other. As humans we can have significant difficulty understanding the clustering of words to each other. To a machine, it represents a huge challenge. To address this problem, this paper describes a new words-clustering technique based on certain text characteristics; by building a system to cluster words in the text depending on characteristics such as morphological, syntactic and Semantic. The clustering is a method of Unsupervised Machine Learning methods, where it collects words with other have similar characteristics in the clusters based on Similarity Function to calculate the distance between those words. We depended on k-mean Clustering to calculate the distance between words, then generating clusters for all referred words in the text. Finally, we will evaluate our system results by using the common evaluation methods. There are three methods of evaluation methods: Precision, Recall, and F-Measure.

In this paper, an adaptive hierarchical clustering method based on DBSCAN algorithm is proposed to get information better from Automatic Identification System, and to be aware of traffic situation on water scientifically. In order to deal with the uneven distribution of ship trajectories, the paper proposes a method of hierarchical clustering and a statistical method to determine parameters according to the data set itself. After experiments, it is proven that the method can identify ship trajectories by their similarity as clusters, and result conforms to the actual traffic flow, thus, the method can provide assistant decision-making in the field of waterway construction, maritime control and so on.

Shared Nearest Neighbor (SNN) Clustering is a well-established density based clustering algorithm, which can find clusters of different sizes, shapes, and densities. SNN has been widely adopted in numerous applications. As the size of dataset becomes extremely large nowadays, it is inefficient or even impossible for large-scale data to be stored and processed on a single machine. Therefore, the scalability problem of clustering algorithm running on a single machine has to be addressed. In this paper, we improve the traditional SNN clustering algorithm by utilizing high-performance computing clusters and powerful programming platform (MapReduce) for big data analysis. In particular, we design the MapReduce-based Shared Nearest Neighbor clustering algorithm called MR-SNN for big data analysis.

In image classification, feature combination is often used to combine the merits of multiple complementary features and improve the classification accuracy compared with one single feature. Existing feature combination algorithms, e.g., multiple kernel learning, usually determine the weights of features based on the optimization with respect to some classifier-dependent objective function. These algorithms are often computationally expensive, and in some cases are found to perform no better than simple baselines. In this paper, we solve the feature combination problem from a totally different perspective. Our algorithm is based on the simple idea of combining only base kernels suitable to be combined. Since the very aim of feature combination is to obtain the highest possible classification accuracy, we measure the combination suitableness of two base kernels by the maximum possible cross-validation accuracy of their combined kernel. By regarding the pairwise suitableness as the kernel adjacency, we obtain a weighted graph of all base kernels and find that the base kernels suitable to be combined correspond to a cluster in the graph. We then use the dominant sets algorithm to find the cluster and determine the weights of base kernels automatically. In this way, we transform the kernel combination problem into a clustering one. Our algorithm can be implemented in parallel easily and the running time can be adjusted based on available memory to a large extent. In experiments on several data sets, our algorithm generates comparable classification accuracy with the state of the art.

Clustering groups the objects into clusters having similarity with each other. This paper focuses on the two techniques of clustering i.e. hierarchical clustering and k-means clustering. The research is to compare various similarities measuring methods and finding out the best one. Research work is started by selecting different categories of textual contents or articles. For each selected category, articles have been selected from various news channels. Search words are identified which are most relevant for a respective category. Now these words are used as input for processing in the program to create a matrix of words. This matrix is then processed in Matlab using different measuring methods. The final outcome is demonstrated by the Cophenatic correlation coefficient & Silhouette Value to find out the best method of similarity measure. In this paper, five categories have been selected for the analysis which are “Business”, “Education”, “Election”, “Entertainment” and “Game” and 28 news articles have been filtered out for each category from various news channels. Different numbers of words are selected like 35, 49, 25, 30 and 35 against the mentioned categories for the implementation of the proposed technique. The research work finally concludes that for hierarchical clustering — ‘Cityblock’ and for k-means clustering — ‘Correlation’ is the best method however cityblock is at second position in the k-means clustering.

K-means is one of the most fundamental techniques in clustering. It has been applied in many fields, such as image processing and Natural Language Processing. It has good performance in many cases, especially in dealing with large data sets. However, how to choose the initial cluster centers is a hard problem, different choice may cause the clustering results by K-means unstable even get the local optimum. To solve this problem, many methods have be proposed, while these methods only apply in some certain fields and perform disappointed when we use for text documents clustering. In this paper, we designed a novel density K-means algorithm and apply it in the text document clustering. The experimental results show that it performs better than most of the existing methods in Chinese corpus. Furthermore, compared with other algorithms, our algorithm can effectively decrease the iterations.

It is crucial to determine the optimal number of clusters for the clustering quality in cluster analysis. From the standpoint of sample geometry, two concepts, i.e., the sample clustering dispersion degree and the sample clustering synthesis degree, are defined, and a new clustering validity index is designed. Moreover, a method for determining the optimal number of clusters based on an agglomerative hierarchical clustering (AHC) algorithm is proposed. The new index and the method can evaluate the clustering results produced by the AHC and determine the optimal number of clusters for multiple types of datasets, such as linear, manifold, annular, and convex structures. Theoretical research and experimental results indicate the validity and good performance of the proposed index and the method.

The maturity of structured knowledge bases and semantic resources has contributed to the enhancement of document clustering algorithms, that may take advantage of conceptual representations as an alternative for classic bag-of-words models. However, operating in the semantic space is not always the best choice in those domain where the choice of terms also matters. Moreover, users are usually required to provide a valid number of clusters as input, but this parameter is often hard to guess, due to the exploratory nature of the clustering process. To address these limitations, we propose a multi-view co-clustering approach that processes simultaneously the classic document-term matrix and an enhanced document-concept representation of the same collection of documents. Our algorithm has multiple key-features: it finds an arbitrary number of clusters and provides clusters of terms and concepts as easy-to-interpret summaries. We show the effectiveness of our approach in an extensive experimental study involving several corpora with different levels of complexity.

Clustering is one of the most common unsupervised learning tasks in machine learning and data mining. Clustering algorithms have been used in a plethora of applications across several scientific fields. However, there has been limited research in the clustering of point patterns - sets or multi-sets of unordered elements - that are found in numerous applications and data sources. In this paper, we propose two approaches for clustering point patterns. The first is a non-parametric method based on novel distances for sets. The second is a model-based approach, formulated via random finite set theory, and solved by the Expectation-Maximization algorithm. Numerical experiments show that the proposed methods perform well on both simulated and real data., Comment: Preprint: 23rd Int. Conf. Pattern Recognition (ICPR). Cancun, Mexico, December 2016

In the previous chapters, we had a comprehensive survey on background knowledge of this book including temporal data mining, temporal data clustering, and ensemble learning. Now, we are going to present a novel Hidden Markov Model–based ensemble clustering approach. Such approach is designed to solve the problems in finding the intrinsic number of clusters and model initialization sensitivity, which commonly existed in temporal clustering tasks.