MCBÜ Açıkerişim Sistemi :: Browsing by Subject "Clustering algorithms"

Browsing by Subject "Clustering algorithms"

Now showing 1 - 15 of 15

A fuzzy clustering neural networks for motion equations of synchro-drive robot
(Elsevier Ltd, 2010) Aydin S.
Motion equations for synchro-drive robot Nomad 200 are solved by using fuzzy clustering neural networks. The trajectories of the Nomad 200 are assumed to be composed of line segments and curves. The structure of the curves is determined by only two parameters (turn angle and translational velocity in the curve). The curves of the trajectories are found by using artificial neural networks (ANN) and fuzzy C-means clustered (FCM) ANN. In this study a clustering method is used in order to improve the learning and the test performance of the ANN. The FCM algorithm is successfully used in clustering ANN datasets. Thus, the best of training dataset of ANN is achieved and minimum error values are obtained. It is seen that, FCM-ANN models are better than the classic ANN models. © 2010 Elsevier Ltd. All rights reserved.
A Dynamic Distributed Tree Based Tracking Algorithm for Wireless Sensor Networks
(2010) Alaybeyoglu A.; Kantarci A.; Erciyes K.
We propose a dynamic, distributed tree based tracking algorithm for very fast moving targets in wireless sensor networks, with speeds much higher than reported in literature. The aim of our algorithm is to decrease the miss ratio and the energy consumption while tracking objects that move in high speeds. In order to do this, the root node which is determined dynamically in accordance with the node's distance to the target, forms lookahead spanning trees along the predicted direction of the target. As the miss ratio decreases, the usage of recovery mechanisms which are employed to detect a target again that is moving away from the predicted trajectory also decreases. This decrease reduces the energy consumption and increases the network lifetime. We describe all the phases of the algorithm in detail and show by simulations that the proposed algorithm performs well to track very fast moving targets. We also compare the algorithm with the generic cluster, generic tree and dynamic multi cluster based tracking algorithms in terms of miss ratio and energy consumption. © Springer-Verlag Berlin Heidelberg 2010.
Fast global fuzzy C-means clustering for ECG signal classification; [EKG i̇şaretlerini siniflamak için hizli global bulanik C-ortalama öbekleşme]
(2010) Koçyiǧit Y.; Kiliç I.
Fuzzy clustering plays an important role in solving problems in the areas of pattern recognition and fuzzy model identification. The Fuzzy C-Means algorithm is one of widely used algorithms. It is based on optimizing an objective function, being responsive to initial conditions; the algorithm usually leads to local minimum results. Aiming at above problem, the fast global Fuzzy C-Means clustering algorithm (FGFCM) has been proposed, which is an incremental approach to clustering, and does not depend on any initial conditions. The algorithm was applied on ECG signals to classification. ©2010 IEEE.
A distributed wakening based target tracking protocol for wireless sensor networks
(2010) Alaybeyoglu A.; Dagdeviren O.; Kantarci A.; Erciyes K.
We propose a two layer protocol for tracking fast targets in sensor networks. At the lower layer, the Distributed Spanning Tree Algorithm (DSTA) [12] partitions the network into clusters with controllable diameter and constructs a spanning tree backbone of clusterheads rooted at the sink. At the upper layer, we propose a target tracking algorithm which wakes clusters of nodes by using the estimated trajectory beforehand, which is different from existing studies [3] in which target can be detected only when the nodes close to the target are awake. We provide the simulation results and show the effect of fore-waking operation by comparing error and miss ratios of existing approaches with our proposed target tracking algorithm. © 2010 IEEE.
Energy efficient target tracking with particle filtering technique in wireless sensor networks
(2011) Alaybeyoglu A.
In this study, a new target tracking algorithm is proposed for wireless sensor networks. The aim of the algorithm is to decrease energy consumption of the system by decreasing the ratio of target misses. Next location of the target is predicted by using Particle Filtering (PF) technique which aims to represent the posterior density function by a set of random samples with associated weights. Nodes are deployed according to the hexagon shaped network topology in which each of the hexagons represents a cluster with a predetermined leader node. In order to decrease the ratio of target misses, nodes that are closer to the target's predicted location are woken up to make them ready to detect the target. This increases the probability of detecting the target by one of the neighboring hexagons when the target makes sudden turns or unexpected movements. Tracking performance of the proposed algorithm is evaluated by comparing with KNearest Cluster Tracking (KNCT), Wakening Based Target Tracking Algorithm (WBTA)[10] and Generic Static Tracking Approach (GSTA) in terms of miss ratio and energy consumption metrics. © Association for Scientific Research.
An adaptive cone based distributed tracking algorithm for a highly dynamic target in wireless sensor networks
(Inderscience Publishers, 2013) Alaybeyoglu A.; Erciyes K.; Kantarci A.
Accurate tracking of a target is imperative in military as well as civil applications. In this study, we propose a distributed cone based tracking algorithm for a target that can move with highly dynamic kinematics along linear and nonlinear trajectories. The algorithm provides wakening of a group of nodes in a cone shaped region along the trajectory of the target and particle filtering is used in the prediction of the next state of the target to decrease the target missing ratios. Algorithm used is adaptive in that the shape of the cone is determined dynamically in accordance with the target kinematics. We compared our algorithm with traditional tracking approaches, a recent tracking algorithm (Semi-Dynamic Clustering (SDC)) and other tracking algorithms that we have previously proposed. Simulation results show that, in terms of target missing ratios, our algorithm is superior to all of these algorithms. Secondly, lower target missing ratios lead to less frequent execution of recovery mechanisms which in turn results in lower energy consumptions. © 2013 Inderscience Enterprises Ltd.
A dynamic lookahead tree based tracking algorithm for wireless sensor networks using particle filtering technique
(Elsevier Ltd, 2014) Alaybeyoglu A.; Kantarci A.; Erciyes K.
In this study, five different algorithms are provided for tracking targets that move very fast in wireless sensor networks. The first algorithm is static and clusters are formed initially at the time of network deployment. In the second algorithm, clusters that have members at one hop distance from the cluster head are provided dynamically. In the third algorithm, clustered trees where members of a cluster may be more than one hop distance from the cluster head are provided dynamically. In the fourth, algorithm lookahead trees are formed along the predicted trajectory of the target dynamically. Linear, Kalman and particle filtering techniques are used to predict the target's next state. The algorithms are compared for linear and nonlinear motions of the target against tracking accuracy, energy consumption and missing ratio parameters. Simulation results show that, for all cases, better performance results are obtained in the dynamic lookahead tree based tracking approach. © 2013 Elsevier Ltd. All rights reserved.
Integrating multi criteria analysis and clustering techniques for the segmentation of after-sale spare part inventory
(Computers and Industrial Engineering, 2014) Güçdemir H.; Ilgin M.A.
Timely and cost effective supply of spare parts is a vital issue in after sales service. If the demand for spare parts is overestimated, holding costs increase. Underestimation of demand results in lost goodwill of the customers. In order to manage the spare part inventories effectively, companies generally determine the importance level of each spare part and apply a suitable inventory control policy. In this study, we integrate clustering techniques and analytic hierarchy process (AHP) to determine the importance levels of spare parts kept by a television manufacturer. First, spare parts are grouped into 3 main categories using clustering algorithms. Then AHP is used to determine importance level of each group by considering several criteria including frequency, criticality, total monthly usage, lead time, availability, substitutability and tendency of obsolescence. Finally, suitable inventory policies are suggested for each spare part group.
An improved ant algorithm with LDA-based representation for text document clustering
(SAGE Publications Ltd, 2017) Onan A.; Bulut H.; Korukoglu S.
Document clustering can be applied in document organisation and browsing, document summarisation and classification. The identification of an appropriate representation for textual documents is extremely important for the performance of clustering or classification algorithms. Textual documents suffer from the high dimensionality and irrelevancy of text features. Besides, conventional clustering algorithms suffer from several shortcomings, such as slow convergence and sensitivity to the initial value. To tackle the problems of conventional clustering algorithms, metaheuristic algorithms are frequently applied to clustering. In this paper, an improved ant clustering algorithm is presented, where two novel heuristic methods are proposed to enhance the clustering quality of ant-based clustering. In addition, the latent Dirichlet allocation (LDA) is used to represent textual documents in a compact and efficient way. The clustering quality of the proposed ant clustering algorithm is compared to the conventional clustering algorithms using 25 text benchmarks in terms of F-measure values. The experimental results indicate that the proposed clustering scheme outperforms the compared conventional and metaheuristic clustering methods for textual documents. © Chartered Institute of Library and Information Professionals.
Hybrid supervised clustering based ensemble scheme for text classification
(Emerald Group Publishing Ltd., 2017) Onan A.
Purpose: The immense quantity of available unstructured text documents serve as one of the largest source of information. Text classification can be an essential task for many purposes in information retrieval, such as document organization, text filtering and sentiment analysis. Ensemble learning has been extensively studied to construct efficient text classification schemes with higher predictive performance and generalization ability. The purpose of this paper is to provide diversity among the classification algorithms of ensemble, which is a key issue in the ensemble design. Design/methodology/approach: An ensemble scheme based on hybrid supervised clustering is presented for text classification. In the presented scheme, supervised hybrid clustering, which is based on cuckoo search algorithm and k-means, is introduced to partition the data samples of each class into clusters so that training subsets with higher diversities can be provided. Each classifier is trained on the diversified training subsets and the predictions of individual classifiers are combined by the majority voting rule. The predictive performance of the proposed classifier ensemble is compared to conventional classification algorithms (such as Naïve Bayes, logistic regression, support vector machines and C4.5 algorithm) and ensemble learning methods (such as AdaBoost, bagging and random subspace) using 11 text benchmarks. Findings: The experimental results indicate that the presented classifier ensemble outperforms the conventional classification algorithms and ensemble learning methods for text classification. Originality/value: The presented ensemble scheme is the first to use supervised clustering to obtain diverse ensemble for text classification © 2017, © Emerald Publishing Limited.
A K-medoids based clustering scheme with an application to document clustering
(Institute of Electrical and Electronics Engineers Inc., 2017) Onan A.
Clustering is an important unsupervised data analysis technique, which divides data objects into clusters based on similarity. Clustering has been studied and applied in many different fields, including pattern recognition, data mining, decision science and statistics. Clustering algorithms can be mainly classified as hierarchical and partitional clustering approaches. Partitioning around medoids (PAM) is a partitional clustering algorithms, which is less sensitive to outliers, but greatly affected by the poor initialization of medoids. In this paper, we augment the randomized seeding technique to overcome problem of poor initialization of medoids in PAM algorithm. The proposed approach (PAM++) is compared with other partitional clustering algorithms, such as K-means and K-means++ on text document clustering benchmarks and evaluated in terms of F-measure. The results for experiments indicate that the randomized seeding can improve the performance of PAM algorithm on text document clustering. © 2017 IEEE.
A hybrid ensemble pruning approach based on consensus clustering and multi-objective evolutionary algorithm for sentiment classification
(Elsevier Ltd, 2017) Onan A.; Korukoğlu S.; Bulut H.
Sentiment analysis is a critical task of extracting subjective information from online text documents. Ensemble learning can be employed to obtain more robust classification schemes. However, most approaches in the field incorporated feature engineering to build efficient sentiment classifiers. The purpose of our research is to establish an effective sentiment classification scheme by pursuing the paradigm of ensemble pruning. Ensemble pruning is a crucial method to build classifier ensembles with high predictive accuracy and efficiency. Previous studies employed exponential search, randomized search, sequential search, ranking based pruning and clustering based pruning. However, there are tradeoffs in selecting the ensemble pruning methods. In this regard, hybrid ensemble pruning schemes can be more promising. In this study, we propose a hybrid ensemble pruning scheme based on clustering and randomized search for text sentiment classification. Furthermore, a consensus clustering scheme is presented to deal with the instability of clustering results. The classifiers of the ensemble are initially clustered into groups according to their predictive characteristics. Then, two classifiers from each cluster are selected as candidate classifiers based on their pairwise diversity. The search space of candidate classifiers is explored by the elitist Pareto-based multi-objective evolutionary algorithm. For the evaluation task, the proposed scheme is tested on twelve balanced and unbalanced benchmark text classification tasks. In addition, the proposed approach is experimentally compared with three ensemble methods (AdaBoost, Bagging and Random Subspace) and three ensemble pruning algorithms (ensemble selection from libraries of models, Bagging ensemble selection and LibD3C algorithm). Results demonstrate that the consensus clustering and the elitist pareto-based multi-objective evolutionary algorithm can be effectively used in ensemble pruning. The experimental analysis with conventional ensemble methods and pruning algorithms indicates the validity and effectiveness of the proposed scheme. © 2017 Elsevier Ltd
Biomedical Text Categorization Based on Ensemble Pruning and Optimized Topic Modelling
(Hindawi Limited, 2018) Onan A.
Text mining is an important research direction, which involves several fields, such as information retrieval, information extraction, and text categorization. In this paper, we propose an efficient multiple classifier approach to text categorization based on swarm-optimized topic modelling. The Latent Dirichlet allocation (LDA) can overcome the high dimensionality problem of vector space model, but identifying appropriate parameter values is critical to performance of LDA. Swarm-optimized approach estimates the parameters of LDA, including the number of topics and all the other parameters involved in LDA. The hybrid ensemble pruning approach based on combined diversity measures and clustering aims to obtain a multiple classifier system with high predictive performance and better diversity. In this scheme, four different diversity measures (namely, disagreement measure, Q-statistics, the correlation coefficient, and the double fault measure) among classifiers of the ensemble are combined. Based on the combined diversity matrix, a swarm intelligence based clustering algorithm is employed to partition the classifiers into a number of disjoint groups and one classifier (with the highest predictive performance) from each cluster is selected to build the final multiple classifier system. The experimental results based on five biomedical text benchmarks have been conducted. In the swarm-optimized LDA, different metaheuristic algorithms (such as genetic algorithms, particle swarm optimization, firefly algorithm, cuckoo search algorithm, and bat algorithm) are considered. In the ensemble pruning, five metaheuristic clustering algorithms are evaluated. The experimental results on biomedical text benchmarks indicate that swarm-optimized LDA yields better predictive performance compared to the conventional LDA. In addition, the proposed multiple classifier system outperforms the conventional classification algorithms, ensemble learning, and ensemble pruning methods. © 2018 Aytuǧ Onan.
Unsupervised Machine Learning for Fire Resistance Analysis
(Springer Science and Business Media Deutschland GmbH, 2023) Çiftçioğlu A.Ö.; Naser M.Z.
Due to its inert nature, concrete has good fire resistance properties. As such, concrete has often been favored for construction – especially where fire hazard is expected. However, this does not mean that reinforced concrete cannot catch fire. It can still be affected by heat and, if exposed to high temperatures, can eventually break down. Therefore, the fire resistance of the reinforced concrete (RC) columns is a critical concern. There are many ways for assessing the fire resistance of structures, but it is difficult to quantify the fire resistance in quantitative terms. The purpose of this work is to investigate the use of unsupervised machine learning by means of clustering to examine the fire resistance of RC columns. A database of over 144 RC columns subjected to standard fire conditions has been collected and then examined via the interpretable Fuzzy C-Means algorithm (FCM) and the Classification and Regression Tree (CART) model. Our results indicate that this clustering technique groups RC columns into four natural groups – each with specific properties and characteristics. Moreover, the CART model is used to analyze the variables used as the basis for the clustering of RC columns. Accordingly, when RC columns are separated into four natural clusters, the first split occurs due to restrictions, and the second separation is controlled by the compressive strength and reinforcement ratios of the columns. This research might be the first to attempt to leverage clustering analysis to investigate the fire response of RC columns. The findings of the study clearly show that unsupervised machine learning can provide valuable insights to fire engineers often missing from traditional supervised learning. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.
The Comparison of Classical and Bayesian Structural Equation Models Through Ordered Categorical Data: A Case Study of Banking Service Quality
(Gazi Universitesi, 2023) Erkan G.; Dogan M.; Tatlidil H.
This study aims to compare classical Structural Equation Modeling (SEM) and Bayesian Structural Equation Modeling (BSEM) in terms of ordered categorical data. In order to show the relationship between service dimensions and banks’ customers’ satisfactions, a data were analyzed with classical SEM and BSEM parameter estimation methods. In the Banking Service Quality Scale (SERVQUAL), which consists of sequential categorical data, classical SEM and BSEM were compared to evaluate customer satisfaction. In classical SEM, parameter estimations were made according to the Maximum Likelihood (ML) estimation method. In most of the studies using SERVQUAL in the literature, the results found in previous studies could not be used as prior informative because the service dimensions consisted of different number of factors. For this reason, considering that the results could yield similar results with the ML estimation method due to the high sample size, the use of conjugate prior was preferred instead of the non-informative prior due to the ordinal categorical nature of the data in the BSEM analysis. Since the questionnaire used in the study had a Likert type scale structure, the threshold values were calculated for ordered categorical data and used as prior informative. Thus, by using the threshold values obtained from the data set, a faster convergence of the parameters was achieved. As a result, service dimensions affecting satisfaction according to the ML parameter estimation method were found, Assurance, Physical Appearance, and Accessibility. In addition to these, Reliability as a service dimension was found to be also statistically significant in BSEM. © 2023, Gazi Universitesi. All rights reserved.

Browsing by Subject "Clustering algorithms"

Results Per Page

Sort Options