Search

This paper describes our submissions for SemEval-2018 Task 8: Semantic Extraction from CybersecUrity REports using NLP. The DM_NLP participated in two subtasks: SubTask 1 classifies if a sentence is useful for inferring malware actions and capabilities, and SubTask 2 predicts token labels (“Action”, “Entity”, “Modifier” and “Others”) for a given malware-related sentence. Since we leverage results of Subtask 2 directly to infer the result of Subtask 1, the paper focus on the system solving Subtask 2. By taking Subtask 2 as a sequence labeling task, our system relies on a recurrent neural network named BiLSTM-CNN-CRF with rich linguistic features, such as POS tags, dependency parsing labels, chunking labels, NER labels, Brown clustering. Our system achieved the highest F1 score in both token level and phrase level.

We have developed a procedure for hierarchical agglomerative clustering of time series data. To measure the dissimilarity between these data, we use classically the Euclidean distance or we apply the costs of the series nonlinear alignment (time warping). In the latter approach, we use the classical costs or the modified ones. The modification consists in matching short signal segments instead of single signal samples. The procedure is applied to a few datasets from the internet archive of time series. In this archive, the series of the same classes possess visual similarity but their time evolution is often different (the characteristic waves have different location within the individual signals). Therefore the use of the Euclidean distance as the dissimilarity measure gives poor results. After time warping, the nonlinearly aligned signals match each other better, and therefore the total cost of the alignment appears to be a much more effective measure. It results in higher values of the Purity index used to evaluate the results of clustering. In most cases, the proposed modification of the alignment costs definition leads to still higher values of the index. © 2018, Springer International Publishing AG. Ministry of Higher Education: BK-220/RAu-3/2016, BKM-508/RAu-3/2016, POIG.02.03.01-24-099/ Acknowledgements. This work was partially supported by the Ministry of Science and Higher Education funding for statutory activities (BK-220/RAu-3/2016) and the Ministry of Science and Higher Education funding for statutory activities of young researchers (BKM-508/RAu-3/2016). The work was performed using the infrastructure supported by POIG.02.03.01-24-099/13 grant: GeCONiI—Upper Silesian Center for Computational Science and Engineering.

We address part-of-speech (POS) induction by maximizing the mutual information between the induced label and its context. We focus on two training objectives that are amenable to stochastic gradient descent (SGD): a novel generalization of the classical Brown clustering objective and a recently proposed variational lower bound. While both objectives are subject to noise in gradient updates, we show through analysis and experiments that the variational lower bound is robust whereas the generalized Brown objective is vulnerable. We obtain competitive performance on a multitude of datasets and languages with a simple architecture that encodes morphology and context., Comment: NAACL 2019

This paper presents the task of deeply analyzing user requests: the situation in ordering bots where users input an utterance, the bots would hopefully extract its full product descriptions and then parse them to recognize each product information (PI). This information is useful to help bots better understand user requests, and act upon a much wider range of actions. We model it as a two-layer sequence labeling problem and apply CRFs to solve the task. We investigate two different feature settings, which are manually designed and automatically learnt from neural models of LSTM and CNN, to build good CRF models. In designing features, we propose additional ones based on Brown clustering to enhance the performance of CRF models. To verify the effectiveness, we build a corpus in the retail domain to conduct extensive experiments. The results show that automatically learnt features are very effective and commonly yield better performance than manually designed features. In both settings, adding the information of tags in one layer can also boost the performance of the other layer. Overall, we achieve the best performance with the F-measure of 93.08% in recognizing full product descriptions, and the F-measure of 92.97% in recognizing PI. To our knowledge, this is the first attempt towards understanding user utterances in the context of building Vietnamese ordering bots.

The paper addresses clustering problems from combinatorial viewpoints. A systemic survey is presented. The list of considered issues involves the following: (1) literature analysis of basic combinatorial methods and clustering of very large data sets/networks; (2) quality characteristics of clustering solutions; (3) multicriteria clustering models; (4) graph based clustering methods (minimum spanning tree based clustering methods, clique based clustering as detection of cliques/quasi-cliques, correlation clustering, detection of network communities); and (5) fast clustering approaches. Mainly, the presented material is targeted to networking. Numerical examples illustrate models, methods and applications.

Data clustering is one of the most widely used data mining techniques, classifying similar data items into groups on the basis of similarity among the data items. Different issues have been observed while achieving the classification of data into the most suitable grouping. Efficiency of the clustering techniques and accuracy of the resulting groups are two of the main issues. To tackle these issues, recently, optimization based techniques have been used, resulting in enhanced quality of the output and improved efficiency of the clustering process. Swarm Intelligence (SI) is one such technique whose different algorithms have been found effective for this purpose. Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) are the two most prominent SI based techniques. In this paper we analyze the use of PSO for data clustering in particular for clustering in a hierarchical manner. We chose PSO based hierarchical techniques, Evolutionary PSO for clustering (EPSO-clustering) and Hierarchical PSO for clustering (HPSO-clustering). Both these techniques work in a hierarchical agglomerative manner, with HPSO-clustering an extension of EPSO-clustering. It combines the properties of hierarchical and partitional clustering and adds SI based optimization to the process. We evaluate our proposed clustering techniques on different benchmark datasets from UCI machine learning data repository as well as real data that we collected locally from a web server. We used inter-cluster and intra-cluster distances, and execution time to measure the performance of our proposed techniques. For evaluation we selected different clustering techniques that were previously used as benchmarks such as k-means, PSO-clustering, Hierarchical Agglomerative Clustering (HAC) and DBSCAN. The results verify that the proposed techniques perform better on the suggested measures against the benchmarks mentioned.

Deciding a rational number of clusters for the problems to be solved and determining the parameters associated with the clustering algorithms are two critical issues in the configuration of data clustering. Usually, a manual trial-and-error manner is used to induce a feasible configuration. This paper presents an automated clustering method, which determines the clustering configuration automatically. The proposed method is based on the techniques of support-vector machines with a local-search strategy. It starts with running one-class support-vector machines (OCSVM) to partition input data into a random number of clusters. When a result is obtained, the “local-search” mechanism launches several rounds of OCSVM each of which works with a new clustering configuration. Each new configuration is from the current configuration with incrementally modifications. The clustering results obtained from the local searches are post-evaluated by specific clustering validity index and the best one is retained. The clustering configuration of the best result is used by OCSVM for the clustering afterwards. Such a clustering process iterates until no better result can be obtained. This paper describes the clustering algorithm and compares three clustering validity indices, i.e. distance-based index, Davies–Bouldin index, and Xie–Beni index, on their effectiveness. The performance of the proposed method is demonstrated on the segmentation of several aerial images. Experimental results show that the proposed approach is feasible and effective for image segmentation.

Clustering is representative unsupervised classification. Many researchers have proposed clustering algorithms based on mathematical models – methods we call model-based clustering. Clustering techniques are very useful for determining data structures, but model-based clustering is difficult to use for analyzing data correctly because we cannot select a suitable method unless we know the data structure at least partially. The new clustering algorithm we propose introduces soft computing techniques such as fuzzy reasoning in what we call linguistic-based clustering, whose features are not incident to the data structure. We verify the method’s effectiveness through numerical examples.

This paper proposes a subspace clustering algorithm with automatic feature grouping for clustering high-dimensional data. In this algorithm, a new component is introduced into the objective function to capture the feature groups and a new iterative process is defined to optimize the objective function so that the features of high-dimensional data are grouped automatically. Experiments on both synthetic data and real data show that the new algorithm outperforms the FG-k-means algorithm in terms of accuracy and choice of parameters. HighlightsWe study the problem of subspace clustering with feature grouping.We propose a k-means-type algorithm by incorporating feature grouping into the objective function.The algorithm is able to determine feature groups automatically.Experiments on synthetic and real data show that the algorithm performs well.

This research introduces extremely fast and scalable clustering algorithm.The proposed algorithm detects automatically clusters number.Furthermore, this algorithm uses three insensitive tuning parameters. Clustering multi-dense large scale high dimensional numeric datasets is a challenging task duo to high time complexity of most clustering algorithms. Nowadays, data collection tools produce a large amount of data. So, fast algorithms are vital requirement for clustering such data. In this paper, a fast clustering algorithm, called Dimension-based Partitioning and Merging (DPM), is proposed. In DPM, first, data is partitioned into small dense volumes during the successive processing of dataset dimensions. Then, noise is filtered out using dimensional densities of the generated partitions. Finally, merging process is invoked to construct clusters based on partition boundary data samples. DPM algorithm automatically detects the number of data clusters based on three insensitive tuning parameters which decrease the burden of its usage. Performance evaluation of the proposed algorithm using different datasets shows its fastness and accuracy compared to other clustering competitors.

Initialisation of the EM algorithm in model-based clustering is often crucial. Various starting points in the parameter space often lead to different local maxima of the likelihood function and, so to different clustering partitions. Among the several approaches available in the literature, model-based agglomerative hierarchical clustering is used to provide initial partitions in the popular mclust R package. This choice is computationally convenient and often yields good clustering partitions. However, in certain circumstances, poor initial partitions may cause the EM algorithm to converge to a local maximum of the likelihood function. We propose several simple and fast refinements based on data transformations and illustrate them through data examples.

Heterogeneous information networks have drawn much attention in recent years due to their significant applications, such as text mining, e-commerce, social networks, and bioinformatics. Clustering different types of objects simultaneously based upon not only their relations of the same type, but also the relations between different types of objects can improve the clustering quality mutually. In this paper, we propose a general model, in which both the homogeneous and heterogeneous relations are considered simultaneously, to describe the structure of the heterogeneous information networks and devise a novel parametric free multi-type overlapped clustering approach. In this model, different types of relations between different types of objects are represented by a group of matrices. In this way, we transfer the multi-type clustering problem into the information compression problem. Subsequently, greedy search approaches, which aim at describing the group of relational matrices with least bits, are proposed. Moreover, by discovering the discriminative clusters among different types of objects, we devise effective parameter-free strategies to discover either overlapping or non-overlapping structure among different types of clusters. Extensive experiments on real-world and synthetic data sets demonstrate our methods are effective and efficient.

The study identifies a need for efficient and robust visual clustering approach that can potentially deal with complex supply chain clustering problems. Based on the underlying philosophy of group technology, a growing hierarchical self-organising map algorithm (GHSOM) is proposed to identify a lower two-dimension visual clustering map that can effectively address supply chain clustering problems. The proposed approach provides optimal solutions by decomposing a large-sized supply chain problem into independent, small, manageable problems. It facilitates simple decision-making by exploring similar clusters that are represented by the neighbouring branches in the GHSOM map structure. Unlike other approaches in literature, the proposed approach can further attain good topological ordered representations of the various work order families, to be processed by clusters of supply units along with information on hierarchical sub-cell formation as identifiable from the visually navigable map. The proposed approach has been successfully applied on 16 benchmarked problems. The performance of GHSOM based on grouping efficacy measure outperformed the best results in literature.

Data clustering plays an important role in partitioning the large set of data objects into known/unknown number of groups or clusters so that the objects in each cluster are having high degree of similarity while objects in different clusters are dissimilar to each other. Recently a number of data clustering methods are explored by using traditional methods as well as nature inspired swarm intelligence algorithms. In this paper, a new data clustering method using modified bat algorithm is presented. The experimental results show that the proposed algorithm is suitable for data clustering in an efficient and robust way.

Discriminative clustering is an unsupervised learning framework which introduces the discriminative learning rule of supervised classification into clustering. The underlying assumption is that a good partition (clustering) of the data should yield high discrimination, namely, the partitioned data can be easily classified by some classification algorithms. In this paper, we propose three discriminative clustering approaches based on Extreme Learning Machine (ELM). The first algorithm iteratively trains weighted ELM (W-ELM) classifier to gradually maximize the data discrimination. The second and third methods are both built on Fisher's Linear Discriminant Analysis (LDA); but one approach adopts alternative optimization, while the other leverages kernel k -means. We show that the proposed algorithms can be easily implemented, and yield competitive clustering accuracy on real world data sets compared to state-of-the-art clustering methods.

mining is a process that explores and analyses large data sets in order to discover meaningful patterns. Clustering is a main task of exploratory data analysis and data mining applications. Clustering is the task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar to each other than to those in other groups (clusters). Clustering has wide applications in the field of medicine, business and marketing, World Wide Web, computer science, social science, educational data mining, climatology and many more. This paper mainly presents an overview of types of clustering techniques and some of the applications of data mining where clustering techniques can be applied. The main goal of clustering is to produce a good and high quality clusters that depends mainly on the similarity measure which has the ability to discover some or all hidden patterns and also make the analysis of data easy Keywordsmining, Clustering, Clustering Techniques, Applications, K-Means Clustering

Hierarchical agglomerative clustering (HAC) is a clustering method widely used in various disciplines from astronomy to zoology. HAC is useful for discovering hierarchical structure embedded in input data. The cost of executing HAC on large data is typically high, due to the need for maintaining global inter-cluster distance information throughout the execution. To address this issue, we propose a new parallelization scheme for multi-threaded shared-memory machines based on the concept of nearest-neighbor (NN) chains. The proposed multi-threaded algorithm allocates available threads into two groups, one for managing NN chains and the other for updating distance information. In-depth analysis of our approach gives insight into the ideal configuration of threads and theoretical performance bounds. We evaluate our proposed method by testing it with multiple public datasets and comparing its performance with that of several alternatives. In our test, the proposed method completes hierarchical clustering 3.09-51.79 times faster than the alternatives. Our test results also reveal the effects of performance-limiting factors such as starvation in chain growing, overhead incurred from using synchronization locks, and hardware aspects including memory-bandwidth saturation. According to our evaluation, the proposed scheme is effective in improving the HAC algorithm, achieving significant gains over the alternatives in terms of runtime and scalability.

Nearest neighbor search is a key technique used in hierarchical clustering and its computing complexity decides the performance of the hierarchical clustering algorithm. The time complexity of standard agglomerative hierarchical clustering is O(n3), while the time complexity of more advanced hierarchical clustering algorithms (such as nearest neighbor chain, SLINK and CLINK) is O(n2). This paper presents a new nearest neighbor search method called nearest neighbor boundary (NNB), which first divides a large dataset into independent subset and then finds nearest neighbor of each point in subset. When NNB is used, the time complexity of hierarchical clustering can be reduced to O(n log 2n). Based on NNB, we propose a fast hierarchical clustering algorithm called nearest-neighbor boundary clustering (NBC), and the proposed algorithm can be adapted to the parallel and distributed computing framework. The experimental results demonstrate that our algorithm is practical for large datasets.

Collecting paraphrases using a Paraphrase Recognizer. Fuzzy hierarchical clusters are processed by SVM classifier to detect paraphrases.Agglomeration on verbs is by followed by division on nouns.A novel fuzzy grouping strategy has been utilized for merging clusters.Better performance than existing k-means clustering and other approaches such as cosine similarity and FCM clustering.Applying Word Sense Disambiguation has been found to be beneficial. Paraphrase Extraction involves the discovery of equivalent text segments from large corpora and finds application in tasks such as multi-document summarization and document clustering. Semantic similarity identification is a challenging problem which is further compounded by the large size of the corpus. In this paper a two-stage approach which involves clustering followed by Paraphrase Recognition has been proposed for extraction of sentence-level paraphrases from text collections. In order to handle the ambiguity and inherent variability of natural language a fuzzy hierarchical clustering approach which combines agglomeration based on verbs and division on nouns has been used. Sentences within each resultant cluster are then processed by a machine-learning based Paraphrase Recognizer to discover the paraphrases. The two-stage approach has been applied on the Microsoft Research Paraphrase Corpus and a subset of the Microsoft Research Video Description Corpus. The performance has been evaluated against an existing k-means clustering approach as well as cosine-similarity technique and Fuzzy C-Means clustering and the two-stage system has consistently demonstrated better performance.

Finite mixtures provide a powerful tool for modeling heterogeneous data. Model-based clustering is a broadly used grouping technique that assumes the existence of the one-to-one correspondence between clusters and mixture model components. Although there are many directions of active research in the model-based clustering framework, very little attention has been paid to studying the specific nature of detected clustering solutions. In this paper, we develop an approach for assessing the variability in classifications carried out by the Bayes decision rule. The proposed technique allows assessing significance of each assignment made. We also apply the developed instrument for identifying influential observations that have impact on the structure of the detected partitioning. The proposed diagnostic methodology is studied and illustrated on synthetic data and applied to the analysis of three well-known classification datasets.

the amount of text data stored in computer repositories is growing every day, we need more than ever a reliable way to assemble or classify text documents. Clustering can provide a means of introducing some form of organization to the data, which can also serve to highlight significant patterns and trends .Document clustering is used in many fields such as data mining and information retrieval. This paper presents the results of an experimental study of some common document clustering techniques. In particular, we compare the two main approaches of document clustering, agglomerative hierarchical clustering Modified BIRCH and Partitional clustering algorithm K-means. As a result of comparing both algorithms we attempt to establish appropriate clustering technique to generate qualitative clustering of real world document.

This is the author accepted manuscript. The final version is available from IOS Press via http://dx.doi.org/10.3233/IFS-141518

The Survival of the Fittest is a principle which selects the superior and eliminates the inferior in the nature. This principle has been used in many fields, especially in optimization problem-solving. Clustering in data mining community endeavors to discover unknown representations or patterns hidden in datasets. Hierarchical clustering algorithm (HCA) is a method of cluster analysis which searches the optimal distribution of clusters by a hierarchical structure. Strategies for hierarchical clustering generally have two types: agglomerative with a bottom-up procedure and divisive with a top-down procedure. However, most of the clustering approaches have two disadvantages: the use of distance-based measurement and the difficulty of the clusters integration. In this paper, we propose an optimal probabilistic estimation (OPE) approach by exploiting the Survival of the Fittest principle. We devise a hierarchical clustering algorithm (HCA) based on OPE, also called OPE-HCA. The OPE-HCA combines optimization with probability and agglomerative HCA. Experimental results show that the OPE-HCA has the ability of searching and discovering patterns at different description levels and can also obtain better performance than many clustering algorithms according to NMI and clustering accuracy measures.

Display Omitted Document clustering refers to the task of grouping similar documents and segregating dissimilar documents. It is very useful to find meaningful categories from a large corpus. In practice, the task to categorize a corpus is not so easy, since it generally contains huge documents and the document vectors are high dimensional. This paper introduces a hybrid document clustering technique by combining a new hierarchical and the traditional k-means clustering techniques. A distance function is proposed to find the distance between the hierarchical clusters. Initially the algorithm constructs some clusters by the hierarchical clustering technique using the new distance function. Then k-means algorithm is performed by using the centroids of the hierarchical clusters to group the documents that are not included in the hierarchical clusters. The major advantage of the proposed distance function is that it is able to find the nature of the corpora by varying a similarity threshold. Thus the proposed clustering technique does not require the number of clusters prior to executing the algorithm. In this way the initial random selection of k centroids for k-means algorithm is not needed for the proposed method. The experimental evaluation using Reuter, Ohsumed and various TREC data sets shows that the proposed method performs significantly better than several other document clustering techniques. F-measure and normalized mutual information are used to show that the proposed method is effectively grouping the text data sets.

This paper presents a novel hierarchical clustering method using support vector machines. A common approach for hierarchical clustering is to use distance for the task. However, different choices for computing inter-cluster distances often lead to fairly distinct clustering outcomes, causing interpretation difficulties in practice. In this paper, we propose to use a one-class support vector machine (OC-SVM) to directly find high-density regions of data. Our algorithm generates nested set estimates using the OC-SVM and exploits the hierarchical structure of the estimated sets. We demonstrate the proposed algorithm on synthetic datasets. The cluster hierarchy is visualized with dendrograms and spanning trees.

Pattern-based clustering algorithms return a set of patterns that describe the objects of each cluster. The most recent algorithms proposed in this approach extract patterns on numerical datasets by applying an a priori discretization process, which may cause information loss. In this paper, we introduce a new pattern-based clustering algorithm for numerical datasets, which does not need an a priori discretization on numerical features. The new algorithm extracts, from a collection of trees generated through a new induction procedure, a small subset of patterns useful for clustering. Experimental results show that the patterns extracted by the proposed algorithm allows to build a pattern-based clustering algorithm, which obtains better clustering results than recent pattern-based clustering algorithms. In addition, the proposed algorithm obtains similar clustering results, in quality, than traditional clustering algorithms.

Time series clustering is widely applied in various areas. Existing researches focus mainly on distance measures between two time series, such as dynamic time warping (DTW) based methods, edit-distance based methods, and shapelets-based methods. In this work, we experimentally demonstrate, for the first time, that no single distance measure performs significantly better than others on clustering datasets of time series where spectral clustering is used. As such, a question arises as to how to choose an appropriate measure for a given dataset of time series. To answer this question, we propose an integration scheme that incorporates multiple distance measures using semi-supervised clustering. Our approach is able to integrate all the measures by extracting valuable underlying information for the clustering. To the best of our knowledge, this work demonstrates for the first time that the semi-supervised clustering method based on constraints is able to enhance time series clustering by combining multiple distance measures. Having tested on clustering various time series datasets, we show that our method outperforms individual measures, as well as typical integration approaches.

Clustering is an extremely studied in data mining problem for the text mining domains. Difficulty finds in various applications like customer segmentation, visualization, and collaborative filtering, classification, indexing and document organization. In text mining, clustering the sentence is the processes and it is used within a general text mining tasks. Some of the clustering algorithms and methods are used for clustering the documents at sentence level. In text clustering, sentence level clustering plays a important role this is used in text mining activities. The cluster size may change from one cluster to another and traditional clustering algorithms have some problems in clustering while the input in the form of data set. This gives problems such as, instability of clusters, sensitivity and complexity. To overcome the limitations of those clustering algorithms, in this paper proposes a algorithm called Fuzzy Relational Eigenvector Centrality-based Clustering Algorithm (FRECCA) which is used for the clustering of sentences. We can obtain the more efficient method to overcome the problems in these existing approaches.

There are many types of clustering algorithms such as centroid, hierarchical, or density-based methods. Each algorithm has unique data grouping principles, which creates different varieties of clusters. Ordering Points To Identify the Clustering Structure (OPTICS) is a well-known density-based algorithm to analyze arbitrary shaped and varying density clusters, but the obtained clusters only correlate loosely. Hierarchical agglomerative clustering (HAC) reveals a hierarchical structure of clusters, but is unable to clearly find non-convex shaped clusters. In this paper, we provide a novel hierarchy generation framework and application which can aid users by combining the advantages of the two clustering methods.

Model-based clustering is a method that clusters data with an assumption of a statistical model structure. In this paper, we propose a novel model-based hierarchical clustering method for a finite statistical mixture model based on the Fisher distribution. The main foci of the proposed method are: (a) provide efficient solution to estimate the parameters of a Fisher mixture model (FMM); (b) generate a hierarchy of FMMs and (c) select the optimal model. To this aim, we develop a Bregman soft clustering method for FMM. Our model estimation strategy exploits Bregman divergence and hierarchical agglomerative clustering. Whereas, our model selection strategy comprises a parsimony-based approach and an evaluation graph-based approach. We empirically validate our proposed method by applying it on simulated data. Next, we apply the method on real data to perform depth image analysis. We demonstrate that the proposed clustering method can be used as a potential tool for unsupervised depth image analysis.

According to the characteristics of the constrained resource in distributed real-time data mining in the Internet of Things (IOT) environment, a distributed data mining method is researched in such environment. Based on the limits of computing ability, storage ability, battery energy resources, network bandwidth, and the Internet single point failure, the distributed network data mining method is researched, and the adaptive technology and peer-to-peer node method are adopted. The DRA-Kmeans algorithm of data mining based on theK-means algorithm is proposed, and the amount of data communication among the sites to reduce the number of iterations and clustering is reduced. Clustering efficiency is improved, and better clustering results and execution efficiency are achieved.

The evaluation of clustering results is one of the most important issues in cluster analysis, a core task for effective information access. There are two types of measures for evaluating the quality of clustering results: internal and external. External validity measures evaluate how well the clustering results match prior knowledge about the data, whereas internal measures do not need external information, dealing only with information within the data. In this regard, the main drawback of external evaluation measures is that they are not applicable in real-world situations. In this paper we present an experimental study to determine whether it is possible to predict the quality of multilingual news clustering results by means of an internal evaluation measure. We study whether the internal evaluation measure Expected Density correlates with the external measure F-measure, the most common way of evaluating clustering results. In the experiments, we use different data collections, clustering algorithms and similarity measures in order to determine their influence in the correlation between those measures. Regarding similarity measures, another important issue in clustering, we propose a new similarity measure to calculate how similar two news documents are. This measure is based on the Named Entities shared by both documents. The results show that correlation depends on several different factors, such as the type of collection, the granularity of the clusters, the type of algorithm and the similarity measure.

We propose a parallelization scheme for an existing algorithm for constructing a web-directory, that contains categories of web documents organized hierarchically. The clustering algorithm automatically infers the number of clusters using a quality function based on graph cuts. A parallel implementation of the algorithm has been developed to run on a cluster of multi-core processors interconnected by an intranet. The effect of the well-known Latent Semantic Indexing on the performance of the clustering algorithm is also considered. The parallelized graph-cut based clustering algorithm achieves an F-measure in the range [0.69,0.91] for the generated leaf-level clusters while yielding a precision-recall performance in the range [0.66,0.84] for the entire hierarchy of the generated clusters. As measured via empirical observations, the parallel algorithm achieves an average speedup of 7.38 over its sequential variant, at the same time yielding a better clustering performance than the sequential algorithm in terms of F-measure. Copyright © 2015 John Wiley & Sons, Ltd.

through the continuously rising amounts of data in technical, medical, scientific, engineering, industrial and monetary fields and their renovation to logical form for the human user is one of the main requirements. To quickly discover and analyze complex patterns and requirements, we need the efficient techniques and need to learn from new data will be necessary for information-intensive applications. One of the solutions for this is that classification and clustering of largely available data. To partially fulfill the industry requirement, in this paper we proposed a two-level approach for clustering large data set for rain fall data prediction with Self Organized Maps (SOM) and Support Vector Machine (SVM) with ID3. In this paper, a novel approach to clustering of the SOM and SVM with ID3 are considered. In particular, the use of hierarchical agglomerative clustering and partitioned clustering with ID3 are investigated. The two- stage procedure first using SOM to produce the prototypes and later it considers the SVM with ID3, that are then clustered in the second stage is found to perform well when compared with direct clustering of the data and to reduce the computation time. KeywordsSVM, ID3, Clustering, Classification, Data mining.

Data clustering aims to segment a database into groups of objects based on the similarity among these objects. Due to its unsupervised nature, the search for a good-quality solution can become a complex process. There is currently a wide range of clustering algorithms, and selecting the best one for a given problem can be a slow and costly process. In 1976, Rice formulated the Algorithm Selection Problem (ASP), which postulates that the algorithm performance can be predicted based on the structural characteristics of the problem. Meta-learning brings the concept of learning about learning; that is, the meta-knowledge obtained from the algorithm learning process allows the improvement of the algorithm performance. Meta-learning has a major intersection with data mining in classification problems, in which it is normally used to recommend algorithms. The present paper proposes new ways to obtain meta-knowledge for clustering tasks. Specifically, two contributions are explored here: (1) a new approach to characterize clustering problems based on the similarity among objects; and (2) new methods to combine internal indices for ranking algorithms based on their performance on the problems. Experiments were conducted to evaluate the recommendation quality. The results show that the new meta-knowledge provides high-quality algorithm selection for clustering tasks.

Support-based clustering using kernels suffers from serious computational limitations inherent in many kernel methods when applied to very large-scale problems despite its ability to identify clusters with complex shapes. In this paper, we propose a novel clustering algorithm called Voronoi cell-based clustering to expedite support-based clustering using kernels. In contrast to previous studies, including the basin cell-based method, the proposed method achieves computational efficiency in both the training phase to construct a support estimate using sampled data to reduce the evaluation of kernels and the labeling phase to assign a cluster label on each data point nearest its representative point. The performance superiority of the proposed method over the other basin cell-based methods in terms of computational time and storage efficiency is verified by various experiments using benchmark sets and in real applications to image segmentation.

An efficient hybrid hierarchical clustering is proposed based on agglomerative method.It performs consistently with different distance measures.It performs consistently on data with different distributions and sizes. Hierarchical clustering is of great importance in data analytics especially because of the exponential growth of real-world data. Often these data are unlabelled and there is little prior domain knowledge available. One challenge in handling these huge data collections is the computational cost. In this paper, we aim to improve the efficiency by introducing a set of methods of agglomerative hierarchical clustering. Instead of building cluster hierarchies based on raw data points, our approach builds a hierarchy based on a group of centroids. These centroids represent a group of adjacent points in the data space. By this approach, feature extraction or dimensionality reduction is not required. To evaluate our approach, we have conducted a comprehensive experimental study. We tested the approach with different clustering methods (i.e., UPGMA and SLINK), data distributions, (i.e., normal and uniform), and distance measures (i.e., Euclidean and Canberra). The experimental results indicate that, using the centroid based approach, computational cost can be significantly reduced without compromising the clustering performance. The performance of this approach is relatively consistent regardless the variation of the settings, i.e., clustering methods, data distributions, and distance measures.

Holistic schema matching is the process of carrying off several number of schemas as an input and outputs the correspondences among them. Treating large number of schemas may consume longer time with poor quality. Therefore, several clustering approaches have been proposed in order to reduce the search space by partitioning the data into smaller portions which can facilitate the matching process. However, there is still a demand for improving the partitioning mechanism by avoiding the random initial solutions (centroids) re-sulted from the clustering process. Such random solutions have a significant impact on the matching results. This study aims to integrate correlation clustering and agglomerative hierarchical clustering toward improving the effectiveness of holistic schema matching. The proposed integrated method avoids the random initial so-lutions and the predefined number of centroids. Several preprocessing steps have been performed with using auxiliary information (domain dictionary). The experiments have been carried out on Airfare, Auto and Book datasets from UIUC Web Integration Repository. The proposed method has been compared with K-means and K-medoids clustering methods. As a results the proposed method has outperformed K-means and K-medoids by achieving 0.9, 0.93 and 0.9 of accuracy for Airfare, Auto and Book respectively.

Clustering methods divide a set of observations into groups in such a way that members of the same group are more similar to one another than to the members of the other groups. One of the scientifically well known methods of clustering is the hierarchical agglomerative one. For data of different properties different clustering methods appear favorable. If the data possess locally linear form, application of planar (or hyperplanar) prototypes should be advantageous. However, although a clustering method using planar prototypes, based on a criterion minimization, is known, it has a crucial drawback. It is an infinite extent of such prototypes that can result in addition of very distant data points to a cluster. Such distant points can considerably differ from the majority within a cluster. The goal of this work is to overcome this problem by developing a hierarchical agglomerative clustering method that uses the prototypes confined to the segments of hyperplanes. In the experimental part, we show that for data that possess locally linear form this method is highly competitive to the method of the switching regression models (the accuracy improvement of 24%) as well as to other well-known clustering methods (the accuracy improvement of 16%).

In most practical applications of image retrieval, high-dimensional feature vectors are required, but current multi-dimensional indexing structures lose their efficiency with growth of dimensions. Our goal is to propose a divisive hierarchical clustering-based multi-dimensional indexing structure which is efficient in high-dimensional feature spaces. A projection pursuit method has been used for finding a component of the data, which data's projections onto it maximizes the approximation of negentropy for preparing essential information in order to partitioning of the data space. Various tests and experimental results on high-dimensional datasets indicate the performance of proposed method in comparison with others.

Propose a grid-based clustering algorithm, which is efficient O(N?log??N).Take the idea of region growing in image processing into the clustering algorithm.Significantly faster than k-means variants, such as litekmeans and PRS, spectral clustering, DBSCAN and Greedy EM.The clustering results are comparable to the other clustering algorithms.It is very suitable for real geo-spatial data sets. Geo-spatial data with geographical information explodes as the development of GPS-devices. The data contains certain patterns of users. To dig out the patterns behind the data efficiently, a grid-growing clustering algorithm is introduced. The proposed algorithm takes use of a grid structure, and a novel clustering operation is presented, which considers a grid growing method on the grid structure. The grid structure brings the benefit of efficiency. For large geo-spatial data, the algorithm has competitive strength on the running time. The total time complexity of the algorithm is O(N?log??N), where the time complexity mainly comes from the seed selection step. The grid-growing clustering algorithm is useful when the number of clusters is unknown since the algorithm requires no parameter on the number of clusters. The clusters detected could have arbitrary shapes. Furthermore, sparse areas are treated as outliers/noises in the algorithm. An empirical study on several data sets indicates that the proposed algorithm works much more efficiently than other popular clustering algorithms.

An algorithm for optimizing data clustering in feature space is studied in this work. Using graph Laplacian and extreme learning machine (ELM) mapping technique, we develop an optimal weight matrix W for feature mapping. This work explicitly performs a mapping of the original data for clustering into an optimal feature space, which can further increase the separability of original data in the feature space, and the patterns points in same cluster are still closely clustered. Our method, which can be easily implemented, gets better clustering results than some popular clustering algorithms, like k-means on the original data, kernel clustering method, spectral clustering method, and ELM k-means on data include three UCI real data benchmarks (IRIS data, Wisconsin breast cancer database, and Wine database).

An approach to incorporate users' experience into consensus clustering is proposed.The approach relies on interactive feature selection from textual data.We model an additional (high-level) text representation using the selected features.We explore high-level features to improve the consensus clustering accuracy.Our approach is competitive even when only few features are selected by the users. Consensus clustering and interactive feature selection are very useful methods to extract and manage knowledge from texts. While consensus clustering allows the aggregation of different clustering solutions into a single robust clustering solution, the interactive feature selection facilitates the incorporation of the users' experience in the clustering tasks by selecting a set of textual features, i.e., including user's supervision at the term-level. We propose an approach for incorporating interactive textual feature selection into consensus clustering. Experimental results on several text collections demonstrate that our approach significantly improves consensus clustering accuracy, even when only few textual features are selected by the users.

Graphical abstractDisplay Omitted HighlightsThe proposed approach offers more detailed insight into the cluster's structure and the underlying decision making process.We select different features to describe different objects.We evaluate the similarity between the objects based on the descriptions of objects. In the framework of Axiomatic Fuzzy Set (AFS) theory, we propose a new approach to data clustering. The objective of this clustering is to adhere to some principles of grouping exercised by humans when determining a structure in data. Compared with other clustering approaches, the proposed approach offers more detailed insight into the cluster's structure and the underlying decision making process. This contributes to the enhanced interpretability of the results via the representation capabilities of AFS theory. The effectiveness of the proposed approach is demonstrated by using real-world data, and the obtained results show that the performance of the clustering is comparable with other fuzzy rule-based clustering methods, and benchmark fuzzy clustering methods FCM and K-means. Experimental studies have shown that the proposed fuzzy clustering method can discover the clusters in the data and help specify them in terms of some comprehensive fuzzy rules.

A text clustering algorithm is proposed to overcome the drawback of division based clustering method on sensitivity of estimated class number. Complex features including synonym and co-occurring words are extracted to make a feature space containing more semantic information. Then the divide and merge strategy helps the iteration converge to a reasonable cluster number. Experimental results showed that the dynamically updated center number prevent the deterioration of clustering result when k deviates from the real class numbers. When k is too small or large, the difference of clustering results between FC-DM and k-means is more obvious and FC-DM also outperformed other benchmark algorithms.