Search

An increasingly common practice among retailers is the creation of their own brands. The so-called private labels allow them to compete on price to attract new customers. The goal is to offer a quality similar to that of the traditional brands but at a lower price range. In this research, clustering algorithms based on machine learning techniques are applied to carry out a behavioral customer segmentation for private labels. The research has been completed in four stages: analysis, design, development, and discussion. During the analysis, 1.073 customer loyalty surveys are preprocessed and analyzed. During the design, 23 questions are selected to design experiments. The clustering algorithms used in the investigation are Simple K-Means Algorithm (SKMA) and Expectation-Maximization Clustering (EMC). Experimental results reveal clear differences in the performance of the selected algorithms. For example, when 4 clusters are predefined, SKMA distributes the instances according to the following proportions: 43%, 26%, 15%, and 15%. EMC, instead, distributes the instances in 18%, 23%, 31%, and 29%. In conclusion, the results show that both algorithms, SKMA and EMC, are effective and useful to segment customers based on their preferences. However, the peculiarities of their clustering strategies gives rise to significantly different clusters. [ABSTRACT FROM AUTHOR]

Today modern applications generate massive volume of data and this data increased in tremendous speed. Organizations and enterprises are dealing with this huge amount of data. Today big challenge is accurately analyze, extract and process data in timely manner. To analyzing big data clustering algorithms play important role. The aim of this paper is to provide a self-contained review of the concepts and algorithms related to clustering, as well as providing a comparison and challenges of algorithms in large data set. During choosing the clustering algorithms practitioners are confusing, this will be helpful to them to choose the right algorithms against specific problem. [ABSTRACT FROM AUTHOR]

Recent technologies are an opportunity for creating advanced data analytics tools for various domains. There are diverse methods based on smart technologies for implementing and supporting data analytics in educational environments, e.g. descriptive methods (reports), diagnostic methods (automated analysis), predictive and prescriptive methods (recommendations, alerts). The educational data come from online surveys, biometric sources, but, mostly, from activities in learning management systems (LMS). The current paper presents a new learning analytics instrument - LeProVE - uCluster for Moodle LMS, which creates virtual profiles of the students and aggregates those profiles into clusters which are further used by teachers to optimize the teaching process. The tool is suitable for today's social and cultural values, as the students' performances (reflected in grades, completed assignments) are not analysed only based on a standard a priori settled by the teacher, but considering the colleagues' performances and social interactions as well. Thus, the teacher can use the results obtained by using LeProVE - uCluster to adjust the homework/exams difficulty to the class context and can recommend further useful resources to individual students and clusters of students. The technologies and methods which are used to implement our instrument are modern: Scala, Spark - a general engine for big data processing, Angular JS, data mining algorithms (K-Means) and Big Data Sampling algorithms. To validate the functionality of the proposed tool, we used 2167 anonymised students' profiles which were imported from a Moodle data set freely available for research purposes. Possible interpretations of data and directions to be followed by teachers to improve their teaching skills are also offered. [ABSTRACT FROM AUTHOR]

Cluster analysis is used to identify dissimilar subgroups of objects out of a set of objects based on a combination of rules. In the light of cluster analysis, it is possible to treat dissimilar individuals in an appropriate manner by taking their dissimilarity into consideration. This will be resulted in enhancing the accuracy and efficiency of estimation and prediction models. This study aims to evaluate the performance of different partitioning methods namely, k-means, k-medoids (PAM) and fuzzy and hierarchical methods namely, agglomerative nesting and divisive analysis in grouping the economic events affecting the foreign exchange market. Cluster analysis performed on economic indicators data set depicts the structure of clusters resulted from all algorithms are the same except the single linkage of agglomerative nesting. Poor quality of the clustering structure formed by the single linkage method is confirmed by the lower value of average silhouette width. Comparatively high value of agglomerative coefficient associated with the ward's method reveals the better performance of clustering compared to other linkages. Economic indicators under study are found to be clustered in three groups as performing high, moderate and low impact on the movements of exchange rates. High impact of economic indicators on the exchange rates is reflected by the high volatility at release time and shorter prevailing time of the impact after the release. [ABSTRACT FROM AUTHOR]

It is a promising vision to utilize white spaces, i.e., vacant VHF and UHF TV channels, to satisfy skyrocketing wireless data demand in both outdoor and indoor scenarios. While most prior works have focused on exploring outdoor white spaces, the indoor story is largely open for investigation. Motivated by this observation and that 70% of the spectrum demand comes from indoor environments, we carry out a comprehensive study of exploring indoor white spaces. We first present a large-scale measurement of outdoor and indoor TV spectrum occupancy in 30+ diverse locations in a typical metropolis Hong Kong. Our measurement results confirm abundant white spaces available for exploration in a wide range of areas in metropolises. In particular, more than 50% and 70% of the TV spectrum are white spaces in outdoor and indoor scenarios, respectively. While there are substantially more white spaces in indoor scenarios than in outdoor scenarios, there is no effective solution for identifying indoor white spaces. To fill in this gap, we propose the first system WISER (for White-space Indoor Spectrum EnhanceR), to identify and track indoor white spaces in a building, without requiring user devices to sense the spectrum. We discuss the design space of such system and justify our design choices using intensive real-world measurements. We design the architecture and algorithms to address the inherent challenges. We build a WISER prototype and carry out real-world experiments to evaluate its performance. Our results show that WISER can identify 30%-50% more indoor white spaces with negligible false alarms, as compared to alternative baseline approaches. [ABSTRACT FROM AUTHOR]

Dealing with big amounts of data is one of the challenges for clustering, which causes the need for distribution of large data sets in separate repositories. However, most clustering techniques require the data to be centralized. One of them, the k-means, has been elected one of the most influential data mining algorithms. Although exact distributed versions of the k-means algorithm have been proposed, the algorithm is still sensitive to the selection of the initial cluster prototypes and requires that the number of clusters be specified in advance. Additionally, distributed versions of clustering algorithms usually requires multiple rounds of data transmission. This work tackles the problem of generating an approximated model for distributed clustering, based on k-means, for scenarios where the number of clusters of the distributed data is unknown and the data transmission rate is low or costly. A collection of algorithms is proposed to combine k-means clustering for each distributed subset of the data with a single round of communication. These algorithms are compared from two perspectives: the theoretical one, through asymptotic complexity analyses, and the experimental one, through a comparative evaluation of results obtained from experiments and statistical tests. [ABSTRACT FROM PUBLISHER]

This paper presents partitioning fuzzy c-means clustering algorithms for interval-valued data based on city-block distaces. These fuzzy c-means clustering algorithms give a fuzzy partition and a prototype for each fuzzy cluster by optimizing an adequacy criterion based on suitable adaptive and non-adaptive city-block distances between vectors of intervals. The adaptive city-block distances change at each algorithm iteration and are different from one fuzzy cluster to another. Experiments with real interval-valued data sets show the usefulness of these fuzzy clustering algorithms. [ABSTRACT FROM PUBLISHER]

This paper gives a relational fuzzy c-medoids clustering algorithm that is able to partition objects taking into account simultaneously several dissimilarity matrices. The aim is to obtain a collaborative role of the different dissimilarity matrices in order to obtain a final consensus partition. These matrices could have been obtained using different sets of variables and dissimilarity functions. This algorithm is designed to give a fuzzy partition and a prototype for each fuzzy cluster as well as to learn a relevance weight for each dissimilarity matrix by optimizing an objective function. These relevance weights change at each algorithm's iteration and are different from one cluster to another. Several examples illustrate the usefulness of the proposed algorithm. [ABSTRACT FROM PUBLISHER]

Novelty detection is a useful ability for learning systems, especially in data stream scenarios, where new concepts can appear, known concepts can disappear and concepts can evolve over time. There are several studies in the literature investigating the use of machine learning classification techniques for novelty detection in data streams. However, there is no consensus regarding how to evaluate the performance of these techniques, particular for multiclass problems. In this study, we propose a new evaluation approach for multiclass data streams novelty detection problems. This approach is able to deal with: i) multiclass problems, ii) confusion matrix with a column representing the unknown examples, iii) confusion matrix that increases over time, iv) unsupervised learning, that generates novelties without an association with the problem classes and v) representation of the evaluation measures over time. We evaluate the performance of the proposed approach by known novelty detection algorithms with artificial and real data sets. [ABSTRACT FROM PUBLISHER]

Mobile ad hoc networks are highly vulnerable to attacks due to infrastructure less environment. The features of MANETs are like open media, dynamic changes in topology, cooperative and collaborative algorithm and no centralized monitoring system. These features make the network more prone to malicious attack. The traditional way of protecting networks with firewalls and encryption software is no longer sufficient and effective for those features. In this paper, we present a new cluster based intrusion detection algorithm that takes care of black hole attacks in a MANET. This proposed algorithm is based on trustworthiness of the nodes in a network. The network is considered to be a layered structured. The nodes are member of a cluster. Each cluster has cluster head that takes care of all the members of its own cluster and communicates with cluster head at layer 2 whenever required. The cluster head at layer 2 communicate with all cluster head at layer 1. The selection of cluster head depends on three parameters battery power, mobility and trust value of a node in a cluster. Periodically the cluster head updation is done according to three parameters. We have implemented our algorithm using the NS simulator. The performance of the proposed Intrusion Detection System (IDS) has also been evaluated in the process. The performance graph shows marked improvement as far as packet dropping is concerned. [ABSTRACT FROM PUBLISHER]

The present paper introduces Subtractive clustering method for identifying the ageing of insulation oil of power transformers. This work handles the problem with complex dimensions to an efficient and accurate dimensional area. Feature selection for clustering is done by extracting values from Spectroscopy absorbance curves of various oil samples. Feature validation is done by designing a fuzzy inference engine; the results obtained from fuzzy inference engine are then compared with practical data to enhance/ensure the validity and found that these results were matched with practical data. These features are now plotted in a space as training data samples and further Subtractive clustering algorithm is applied for different radius values to generate an efficient inference engine. The oil groups viz new age, medium age and old age samples are identified and correspondingly centers have been plotted for each group. Verification for this work is done practically in the laboratory and has been found that the proposed methodology is efficient in diagnosing the ageing of insulation oil. [ABSTRACT FROM PUBLISHER]

This paper proposes a weighted (W) k-prototype (KP) Multi Objective Genetic Algorithm (MOGA) (W — KP — MOGA) that can automatically evolve feature weights (based on importance of features in cluster) and clustering solutions. Here we are hybridizing KP with MOGA. Minimization of Homogeneity (H) and maximization of Separation (S) are two measures of optimization. For comparison purpose we have also implemented KP and KP — MOGA. Testing by different real world data set with different clustering validity indices shows the superiority of W — KP — MOGA. [ABSTRACT FROM PUBLISHER]

In this paper, we consider the issue of distributed computation of tensor decompositions. A central unit observing a global data tensor assigns different data sub-tensors to several computing nodes grouped into clusters. The goal is to distribute the computation of a tensor decomposition across the different computing nodes of the network, which is particularly useful when dealing with large-scale data tensors. However, this is only possible when the data sub-tensors assigned to each computing node in a cluster satisfies minimum conditions for uniqueness. By allowing collaboration between computing nodes in a cluster, we show that average consensus based estimation is useful to yield unique estimates of the factor matrices of each data sub-tensor. Moreover, an essentially unique reconstruction of the global factor matrices at the central unit is possible by allowing the sub-tensors assigned to different clusters to overlap in one mode. The proposed approach may be useful to a number of distributed tensor-based estimation problems in signal processing. [ABSTRACT FROM PUBLISHER]

We apply recent developments in clustering theory of asymmetric networks to study the equilibrium configurations of consensus dynamics in trust networks. We show that reciprocal clustering characterizes the equilibrium opinions of mutual trust dynamics. That is, clusters in the reciprocal dendrogram correspond to different equilibrium opinions of mutual trust consensus for varying trust thresholds. Moreover, for unidirectional trust dynamics, we show that aggregating nonreciprocal clusters into single nodes does not modify reachability of global consensus, thus, simplifying the consensus analysis of large networks. [ABSTRACT FROM PUBLISHER]

Uncovering the hidden subtleties and irregularities of the events in the video sequence, is the key issue for automatic video surveillance. Notice the fact that the occurrence of abnormal events is rare while the frequently occurring events become normal in general human perception. So we have proposed the unsupervised learning algorithm, Proximity (Prx) clustering for abnormality detection in the video sequence. Prx clustering tries to select only the dominant class sample points from the dataset. For each data sample, it also assigns the degree of belongingness to the dominant cluster. The proposed motion features viz. circulation, motion homogeneity, motion orientation and stationarity try to extract important information necessary for abnormality detection. After performing Prx clustering, each sample belongs to dominant cluster with the membership value. When Prx clustering is being performed in the space constructed from the proposed motion features, it helps to improve the abnormality detection performance. Experimental results for clustering performance evaluation on artificial dataset show that the Prx clustering outperforms the other clustering methods, for clustering the single dominant class from the dataset. Abnormality detection experiments show the comparable performance with other methods, in addition it has an advantage of incremental learning that it learns about the new normal events in an unsupervised manner. [ABSTRACT FROM PUBLISHER]

A method is proposed for analysing the geometric and logical structure of pages in a typical single-column book. A Gaussian blur combined with thresholding is used to form connected components which nominally represent words. A bottom-up nearest-neighbour approach is used to find textual lines, and a manually-defined line length parameter is used to remove marginal noise and find the page frame. A state machine is used to group lines and label them according to function. The proposed method is able to correctly segment and label 99.82% of all targeted features in a set of 196 sample pages. Of the sixteen errors encountered in the sample pages, eleven are instances where adjacent lines have been merged together, four are instances where paragraphs have been split in half, and the remaining error was caused by a header element being detected as part of the body text. [ABSTRACT FROM PUBLISHER]

Image Segmentation, a basic technique for many real world applications, has been considered in this paper. The Seeded Region Growing (SRG) algorithm, as the first and probably the simplest region growing algorithm, faces three important problems: the position of seeds, the number of seeds, and region growing strategy. Two new versions of SRG are introduced here to solve the multi seeded region growing problem, and also region growing strategy. Then Particle Swarm Optimization is utilized to solve the localization problem. Experimental results show that the proposed method is successfully applied to gray scale image segmentation. [ABSTRACT FROM PUBLISHER]

Manifold clustering is often used to partition a multiple manifold dataset prior to the application of manifold learning. Thus manifold clustering can be seen as a preprocessing step for eliminating singularities in a dataset before doing dimension reduction. In this paper, we propose an algorithm for resolving singularities prior to dimension reduction. We achieve singularity resolution using algebraic blow ups as motivation. With this type of singularity resolution, we are able to simultaneously perform manifold clustering and learning. The algorithm is based on a simple modification of Isomap which identifies and treats singularities before providing reduced dimensional representations. We demonstrate our algorithm with various examples and apply it to problems in molecular conformation, motion segmentation, and face clustering. [ABSTRACT FROM PUBLISHER]

The purpose of the present work is creating an intelligent system to retrieve desired documents in Marathi language. The system also focuses on providing the personalized documents in Marathi language to the end user based on their interests identified from the browsing history. This paper presents the automatic categorization of Marathi documents and the literature survey of the related work done in automatic categorization of text documents. Several supervised learning techniques are exists for the classification of text documents namely Decision trees, Support Vector machine (SVM), Neural Network, Ada Boost and Naïve Bayes etc. Several clustering techniques are also available for text categorization namely K-means, Suffix Tree Clustering (STC), Semantic Online Hierarchical Clustering (SHOC), Label Induction Grouping Algorithm (LINGO) etc. In the literature survey it is found that vector space model (VSM) gives better result than probabilistic model. This paper presents categorization of the Marathi text documents using Lingo Clustering algorithm based on VSM. The data set consists of 107 Marathi documents of 3 different categories-Tourism, Health Programmes and Maharashtra festivals. The result shows that the performance of the LINGO clustering algorithm is good for categorizing the Marathi text documents. For the Marathi documents overall accuracy of the system is 91.10%. [ABSTRACT FROM PUBLISHER]

Detection of the anomaly activities in the network has been a growing problem, motivating widespread research in the area of automated intrusion detection systems. In the automated intrusion detection systems, classification of n-dimensional vectors of the network traffic is a challenging area. Several research works was already done on this topic. But most of the works were presented to have high detection rates, But with false positives. In this paper, we presented a novel approach to have a high detection rate and very low false positives and false negatives in the classification of network traffic using jaccords coefficient (JC) similarity. The proposed approach is employed on low dimensional space of network traffic profiles with the KDDCUP99 dataset. The experimental study shows that the use of jaccords coefficient similarity clustering on the network traffic profile will increases the detection rate and avoids the false positives in the classification. [ABSTRACT FROM PUBLISHER]

Manifold learning based image clustering models are usually employed at local level to deal with images sampled from nonlinear manifold. Usually, gray level image features are used that are obtained by resizing original images through linear interpolation approach. However, significant image variance information is lost in gray level image features. Clustering models that are based on discriminant analysis can be made more effective in principal component analysis (PCA) space whereas leading projection vectors contain significant image variance information. For optimal clustering performance, we used two-dimensional two-directional PCA technique to extract significant image features. We report clustering performance of Spectral Embedded Clustering (SEC) model using discriminant image features on 6 benchmark image databases. Clustering performance is compared with existing state-of-art clustering approaches. Significant overall performance improvement is observed using proposed discriminant image features over gray level image features. [ABSTRACT FROM PUBLISHER]

In their nature, wireless sensor nodes are resource restrained, such as low power capabilities and small memory. These limitations inevitably affect secure energy-efficient wireless sensor networks protocols in such way that some cryptographic algorithms such as asymmetric algorithms are not convenient to be used in wireless sensor nodes. In this paper, based on perturbation-based efficient confidentiality preserving scheme and Energy-Efficient Secure Pattern Based Data Aggregation for Wireless Sensor Networks scheme, we achieve higher energy-efficiency by using pattern-codes approach to avoid energy wastage due to sending redundant data to cluster-heads. Symmetric cryptographic algorithms such as perturbation algorithm used are more convenient to wireless sensors by providing high data privacy and confidentiality as well. [ABSTRACT FROM PUBLISHER]

Clustering sensor nodes is an effective method for achieving high energy efficiency and flexible data aggregation in wireless sensor networks. However, scheduling all clusters to avoid the inter-cluster interference and meanwhile send all observation to the sink quickly is a difficult problem. In this paper, we use interference graph to model the inter-cluster collision problem and propose an efficient sleep scheduling mechanism in the clustered wireless sensor networks. We firstly divide the clusters into layers, then transform the cluster-tree network into the interference graph of clusters, and thereby divide all clusters on one layer into different independent sets. All clusters in the same independent set can work simultaneously without interference, while all independent sets are scheduled one by one for collision-free. To reduce the total delay of data report in the whole network, we propose three kinds of rules to sort the independent sets. Simulation and analysis show that the degree-based sorting rule works best. The practical experiments also verify this conclusion. [ABSTRACT FROM PUBLISHER]

Wireless sensor networks may contain hundreds or thousands number of sensor nodes, monitoring specified parameters in the environment. These sensor nodes have the ability to communicate with each other or directly to an external base station. Current clustering algorithms provide an effective way to prolong the network lifetime. However, gathering information and communicating with each other is crucial in an emergency situation. To address this problem, the paper proposes an emergency response unequal clustering algorithm for wireless sensor networks. The proposed algorithm has been able to utilize the energy usage of the network with the novel approach in unequal clustering. Furthermore, this addresses the node failures in an emergency situation. The simulation result shows that the proposed algorithm has optimized the energy usage and reduce the communication failure compared to other existing algorithms. [ABSTRACT FROM PUBLISHER]

In a Wireless Sensor Network (WSN) hundreds of tiny sensors with limited resources are accommodated to sense the information from the field. Transfer of gathered information from the sensing field to the base station must be done in proficiently to sustain the network longer. Clustering of sensor nodes is one way to achieve this goal. This paper introduces an Energy Efficient clustered routing protocol based on LEACH-C for WSN. In LEACH-C (Low Energy Adaptive Clustering Hierarchy-Centralized), the cluster heads are selected by the base station randomly. This paper introduces a novel cluster based routing protocol in which, the base station finds the highest energy node among the cluster and mark it as a cluster head for the current time. Thus in the proposed system the energy consumption of various nodes becomes more uniform as compared to LEACH-C. The simulation results indicate that our proposed method leads to efficient transmission of data packets with less energy and therefore increases the network longevity as compared to LEACH-C and LEACH. [ABSTRACT FROM PUBLISHER]

The market of smartphones has been exploding, and taking pictures is a basic, maybe one of the most important functions of a smartphone. In this paper we address the problem of managing a large amount of mobile photos by automatically tagging the photos, so they can be easily browsed or searched later. Unlike other content-based photo tagging approaches, this paper's main contribution is to explore an alternative opportunity of automatic photo tagging using contextual information. Both clustering and similarity-based approaches were studied for photo tagging using context such as date, time, location, environment noise, and human faces. The results show that there are intrinsic connections between contextual information and photo tags, and similarity-based approach outperforms clustering-based tagging significantly. [ABSTRACT FROM PUBLISHER]

Network virtualization is an important technique to allow multiple customized networks to cohabit on a shared substrate network. In order to put the network virtualization into practice, the virtual network embedding (VNE) problem that searching an efficient mapping of a virtual network request onto substrate network resources should be solved primarily. Since the VNE problem is known to be NP-hard, previous work focused on designing heuristic-based algorithms. In this paper, we proposed the nodes clustering and dynamic service balance awareness based virtual network embedding (NC-DSBA) that employs the divide-and-conquer strategy along with the concurrence policy. Besides, the proposed NC-DSBA introduced a dynamic service balance algorithm to assist the embedding and improve the efficiency. The simulation results show that the NC-DSBA reduces the computational complexity, increases the acceptance ratio and revenue-cost ratio, and balances the loads of virtual nodes and links in substrate network in the long run. [ABSTRACT FROM PUBLISHER]

Clustering algorithms are being widely used on biomedical data. They aim to extract important information that can be used to improve life conditions by helping specialized technicians on the decision process. Clustering algorithms based on information theory concepts claim that by using higher order statistic they are able to extract more information from the data and therefore provide much better results. In this work we try to verify this claim by comparing the performance of some entropic clustering algorithms against more conventional ones. Results of the performed experiments are not conclusive but they seem to indicate that this kind of entropic algorithms may provide some improvements when clustering biomedical data. [ABSTRACT FROM PUBLISHER]

Over time, it has been found there is valuable information within the data sets generated into different areas. These large data sets required to be processed with any data mining technique to get the hidden knowledge inside them. Due to nowadays many of data sets are integrated with a big number of instances and they do not have any information that can describe them, is necessary to use data mining methods such as clustering so it can permit to lump together the data according to its characteristics. Although there are algorithms that have good results with small or medium size data sets, they can provide poor results when they work with large data sets. Due to above mentioned in this paper we propose to use different cluster validation methods to determine clustering quality, as its analysis, so at the same time to determine in an empiric way the more reliable rates for working with large data sets. [ABSTRACT FROM PUBLISHER]

Knowledge exaction and text representation are considered as the main concepts concerning organizations nowadays. The estimation of the semantic similarity between words provides a valuable method to enable the understanding of texts. In the field of biomedical domains, using Ontologies have been very effective due to their scalability and efficiency. The problem of extracting knowledge from huge amount of data is recorded as an issue in the medical sector. In this paper, we aim to improve knowledge representation by using MeSH Ontology on medical theses data by analyzing the similarity between the keywords within the theses data and keywords after using the MeSH ontology. As a result, we are able to better discover the commonalities between theses data and hence, improve the accuracy of the similarity estimation which in return improves the scientific research sector. Then, K-means cluster algorithm was applied to get the nearest departments that can work together based on medical ontology. Experimental evaluations using 4, 878 theses data set in the medical sector at Cairo University indicate that the proposed approach yields results that correlate more closely with human assessments than other by using the standard ontology (MeSH). Results show that using ontology correlates better, compared to related works, with the similarity assessments provided by experts in biomedicine. [ABSTRACT FROM PUBLISHER]

In this paper, for zoning a large set of location's data we apply the k-means clustering algorithm. The results were plotted graphically and were satisfactory, so we conclude that the algorithm is useful despite the size of the data, at least for low data dimensions (latitude, longitude). [ABSTRACT FROM PUBLISHER]

The node energy consumption is not balanced inexisting clustering routing protocol of wireless sensor network (WSN).A clustering routing algorithm based on distance probability (CRABDP) is proposed to improve this problem. It is calculated probability selected into head according to the distance with sink and the residual energy of nodes. Cluster heads distribution near sink is more than its distribution far from the base station so as to constitute a heterogeneous hierarchynetwork to realize the energy equilibrium. Performance analysis and simulation experiments show thatthe new algorithm can better balanced energy consumption and improve the utilization rate of energy to prolong network life timecompared with the classical LEACH protocol. [ABSTRACT FROM PUBLISHER]

The aim of this research is to segment spontaneous speech using an unsupervised learning technique. We are especially interested from a machine perception or detection point-of-view, and focus on revealing some structure of prosody in spontaneous speech. The BEA spontaneous speech database is used to develop a speech segmentation system. The spontaneous narratives are annotated manually for intonational phrases (IP) and further divided for phonological phrases (PP). Word level transcription is also provided. For the automatic detection of IPs and embedded PPs, a two-step segmentation method is applied. In the first step, the IPs are detected automatically based on speech energy, spectral centroid and a double-thresholding technique. In the second step, PPs are segmented within the IPs, based on F0, energy and Kullback-Leibler divergence combined with an adaptive thresholding method. The results show that the proposed method can provide good and efficient framework for segmenting Hungarian spontaneous speech, with a performance close to read speech. [ABSTRACT FROM PUBLISHER]

In recent years the MapReduce framework has become one of the most popular parallel computing platform for processing big data. It is frequently used by companies such as Facebook, IBM, and Google. Moreover, more and more universities adopt this approach for their academic purposes. Since the MapReduce computing framework is designed for distributed computing on massive data sets, with the help of new algorithms, it can be a suitable platform for processing graphs with billions of vertices and edges. The st-connectivity, that is, the existence of at least one path between any two vertices s and t, is a fundamental graph theory problem. This paper introduces a new algorithm, which is capable of finding st-connectivity in arbitrary graphs, designed for the MapReduce framework. The paper also defines the data structure, which might be suitable in case of other graph processing algorithms. [ABSTRACT FROM PUBLISHER]

We propose a method of generating different scale-free networks, which has several input parameters in order to adjust the structure, so that they can serve as a basis for computer simulation of real-world phenomena. The topological structure of these networks was studied to determine what kind of networks can be produced and how can we give the appropriate values of parameters to get a desired structure. [ABSTRACT FROM PUBLISHER]

This paper aims to develop a novel framework of multi-cell tracking using intelligent ant system, in which a priori colony distribution block is first proposed to directly place birth ants on relevant pixels of current image through kernel density probability estimate of background; afterwards, a multi-colony reconstruction block is developed to further attract ants towards potential regions according to heuristic histogram similarity and pixel pheromone level with an appropriate evaporation and propagation model; finally, a cell state extraction block is implemented to adaptively determine the number of cells and their individual states. Experiment results on real cell image sequences demonstrate that our algorithm could give a more accurate and robust performance than other methods. [ABSTRACT FROM PUBLISHER]

This paper presents Isarn Dhama word segmentation based on the Isarn Dharma writing system and dictionary. In this study, input text is segmented into sequences of Isarn Dharma Character Clusters (IDCCs). Each IDCC represents a group of inseparable Isarn Dharma characters based on the Isarn Dharma writing system. The sequence of IDCCs will be considered as input in order to look for the most suitable segmentation word from the dictionary using the IDCC longest matching algorithm. Grouping rules were then used to group adjacent remaining IDCCs that do not match an Isarn word in the dictionary. In order to evaluate the efficiency of the proposed technique, Isarn literature, Jataka, legend and Buddha foretell were used as the testing data to test the proposed system; comparing with longest matching and a hybrid of the IDCC longest matching. The experiment results showed that the F-measures are 80.15%, 85.06% and 86.07% for the longest matching, the IDCC longest matching algorithm, and the proposed method, respectively. [ABSTRACT FROM PUBLISHER]

Simulation models have importantly expanded the analysis capabilities in engineering designs. With larger computing power, more variables can be modeled to estimate their effect in ever-larger number of performance measures. Statistical experimental designs, however, are still somewhat focused on the variation of less than about a dozen variables. In this work, an effort to identify strategies to deal with tens of variables is undertaken. The aim is to be able to generate designs capable to estimate full-quadratic models. Several strategies are contrasted: (i) generate designs with random numbers, (ii) use designs already in the literature, and (iii) generate designs under a clustering strategy. The first strategy is an easy way to generate a design. The second strategy does focus on statistical properties, but the designs become somewhat inconvenient to generate when increasing the number of variables. The third strategy is currently being investigated as a possibility to provide a balance between (i) and (ii). [ABSTRACT FROM PUBLISHER]

The economic environmental load dispatch problem in power networks aims at producing electricity at the lowest financial and environmental costs. In this paper, we propose a novel real-time dynamic data driven adaptive multi-scale simulation framework (RT-DDDAMS) for efficient real-time dispatching of electricity. The framework includes 1) a discovery procedure where the network is split into sub-networks and prospective fidelities are identified, 2) an RT-DDDAMS platform involving algorithms for state estimation, fidelity selection, and multi-objective optimization alongside with a system simulation; and 3) databases for storing sub-network topologies, fidelities, and selective measurements. The best compromise load dispatch obtained from this framework is then sent to the considered power network for deployment. The proposed framework is illustrated and validated via a modified IEEE-30 bus test system. The experiments reveal that the proposed framework significantly reduces the computational resource usages needed for the reliable power dispatch without compromising the quality of the solutions. [ABSTRACT FROM PUBLISHER]

In modern cellular systems, to improve the spectral efficiency and the system capacity we use multiple base station (BS) cooperation in downlink cellular networks. This method can highly improve the transmission rate when the user is a cell-edge user. But when the user is a cell-center user, it has minimal improvement on the user capacity, and increases the complexity of the network. As the whole system BS cooperation is hardly achieved, grouping BSs into clusters can realize BSs cooperation. But it still suffers from inter-cluster interference especially for the cluster-edge users. In this paper we consider fractional base station cooperation to deal with the problem. The clusters are formed dynamically for the cell-edge users from the users' perspective to reduce the inter-cluster interference, and are allowed to be overlapped. Accordingly, coordinated precoding scheme is adopted to reduce the intra-cluster interference. Simulations show that the proposed scheme provides impressive average capacity and the user fairness is improved significantly. [ABSTRACT FROM PUBLISHER]

This paper gives an example of a novel simple implementation of the EM algorithm for clustering images. Here we use a simple gray scale color feature to describe an image. When compared to results of other methods using the same simple feature, we found that the proposed method performs well. These comparison results imply that this simple model can be extended to cluster images using more complex features such as texture, shape, and other color descriptors to further improve the precision and recall of the results in order to outperform the existing methods. Further research can prove that this simplified EM algorithm achieves robust classification of unrestricted image domain. [ABSTRACT FROM PUBLISHER]

The Ant Clustering Algorithm (ACA) is a biological inspired data clustering technique, which aimed to cluster and classify the data patterns into different groups. This paper shows how the Ant Clustering Algorithm (ACA) is implemented in clustering and classifying the tropical wood species. As for feature extraction in this research, two feature extractors are selected to extract wood features from wood images, which are Basic Grey Level Aura Matrices (BGLAM) and Statistical Properties of Pores Distribution (SPPD). The ACA algorithm is then been applied in wood data training and testing, and as a result, it is proven that the ACA algorithm can cluster and classify the tropical wood data accurately and effectively. [ABSTRACT FROM PUBLISHER]

Previous works have analyzed the cluster organization of the cat cortical network using both traditional multidimensional scaling methods and evolutionary optimization algorithms. Interestingly, the evolutionary optimization principle of previous works is based on the modularity measure used to find communities in network with global algorithms. In this paper, we deepen this point taking into account different community-detection algorithms. We compare the performances of Net Explorer, a local information dynamics algorithm for detecting communities in networks, with six well-known community detection algorithms: Info map, Hierarchical Info map, Lou vain, Modularity Optimization, Label Propagation and Oslom. The results indicate that Net Explorer is able to detect the four functional clusters where misattributions of some areas are explained by their multimodal function. Results are discussed in terms of misattributions of brain areas to the different clusters emphasizing connections which are explainable (or not) by a cognitive point of view. [ABSTRACT FROM PUBLISHER]

In this paper, we propose a novel clustering algorithm based on Fuzzy C-Means (FCM) and Speeded-Up Robust Feature (SURF) for multispectral image segmentation. In the experiments, color images and multispectral images from IRIS Lab data base, which consists of face images taken along the visible spectrum, have been used to illustrate the performances of the proposed algorithm and to compare its outputs with other algorithms. Results demonstrate that the proposed method outperforms other clustering methods in segmenting color images as well as multispectral images. [ABSTRACT FROM PUBLISHER]

Volunteer Computing (VC) is an approach under grid computing which prepares a platform for computer owners around the globe to donate their resources at internet based VC projects like as SETI@home, Folding@hom. Therefore, global desktop grid is a VC that includes hosts who present their resources to solve some kind of CPU, RAM or Disk spaceintensive problems. At the moment, the common approach used in resource scheduling in VC is mostly based on CPU availability alone. However, we believe that RAM and disk space of the volunteering hosts also play a role in ensuring an optimized assignment of a job to a host, and hence, could lead to an improved resource scheduling performance of the VC. As an initial step to support the above proposal, a formula for normalizing the three different resource information, i.e., CPU, RAM, and disk space, into common representations is developed. In this paper, the normalization formula is presented as well as the formula for calculating the average availability of three resources. Besides, the results from applying the two formulas to a data set extracted from http://fta.inria.fr/res/ are also given. The results show a significant support to the potential of the proposed approach. Impact of this research to the organization will be execution of their tasks' project in suitable cluster of volunteers with specified weighted mean of resources. [ABSTRACT FROM PUBLISHER]

Cluster based routing protocols have significant impact on the energy dissipation and life time of wireless sensor networks (WSN). In this paper Basic Low Energy adaptive clustering hierarchy (LEACH) protocol has been modified with our proposed LEACH-MAE(LEACH-Mobile Average Energy based) protocol to overcome its shortcomings to support mobility along with the new average energy based Cluster Head selection technique. Our simulation in NS2 shows that proposed algorithm improves network life time up to 25 % as well as helps to maintain the equal distribution of energy resource among the sensor nodes. [ABSTRACT FROM PUBLISHER]

In this paper, we proposed a new approach for image clustering to address the adverse effects of noise presented in the images. In particular, the concept of information gain has been incorporated into classical fuzzy c-means (FCM) algorithm in order to develop a robust clustering method. FCM is associated with high sensitivity to noise and produces non-homogenous clustering. To induce robustness to noise, the new clustering technique updates fuzzy membership values and cluster centroids based on information gain. The proposed method produces more homogeneous clustering and its performance can be verified at noisy and noise free images. Experiments have been performed on synthetic, CT liver images and compared with those of classical FCM and one of its robust variants. Moreover, the proposed algorithm has been validated on a data set of 30 carotid artery ultrasound images. Visual inspection of segmented images and clustering quality measures confirm that the proposed approach outperforms other clustering algorithms in comparison. Quantitative measures, in terms of PC and CE, also lead to similar conclusion. Hence, the proposed algorithm is robust to noise and produces homogenous clustering. [ABSTRACT FROM PUBLISHER]

Passive millimeter wave (MMW) imaging can create interpretable imagery of objects concealed under clothing. Unfortunately, low signal to noise ratio and low temperature resolution make automatic analysis of passive MMW images difficult. In this paper, we analyze passive MMW images generated by 8 mm regime MMW. The imaging system is composed of two channels: one with linear horizontal polarization and the other with linear vertical polarization. Both registration between horizontal and vertical polarization images and segmentation of concealed objects are addressed. Registration is performed by geometric feature matching and affine transform, while multi-level segmentation separates the human body region from the background, and concealed objects from the body region, sequentially. Experiments measuring average error probability show that our method separate objects with higher accuracy than the conventional method with a single channel image. [ABSTRACT FROM PUBLISHER]

This article deals with the obstacle detection on a railway crossing (clearance detection). Detection is based on the optical flow estimation and classification of the flow vectors by K means clustering algorithm. For classification of passing vehicles is used optical flow direction determination. The optical flow estimation is based on a modified Lucas-Kanade method. [ABSTRACT FROM PUBLISHER]

A huge amount of information about real-time events are being generated in every second in a running IT-Infrastructure and recorded by the system logs, application logs, as well as the output from the deployed security or management methods, e.g., IDS alerts, firewall logs, scanning reports, etc. To rapidly gather, process, correlate, and analyze the massive event information is a challenging task. High performance security analytics is proposed to address this challenge by which the real-time event information can be normalized, centralized, and correlated to help identify the current running state of the target environment. As an example of next generation Security Information and Event Management (SIEM) platform, Security Analytics Lab (SAL) has been designed and implemented based on the newly emerged In-Memory data management technique, which makes it possible to efficiently organize, access, and process different types of event information through a consistent central storage and interface. In this paper, the multi-core architecture is introduced on the event correlation module of SAL platform by which the correlation tasks can be executed in parallel by different computing resources. The k-means algorithm is implemented as an example of possible event clustering and correlation algorithms. Several experiments are conducted and analyzed to show that the performance of analytics can be significantly improved by applying multi-core architecture in the event correlation procedure. [ABSTRACT FROM PUBLISHER]