Incremental Clustering Research Articles

증분형 K-means 클러스터링 기반 방사형 기저함수 신경회로망 모델 설계

In this study, the design methodology of radial basis function neural networks based on incremental K-means clustering is introduced for learning and processing the big data. If there is a lot of dataset to be trained, general clustering may not learn dataset due to the lack of memory capacity. However, the on-line processing of big data could be effectively realized through the parameters operation of recursive least square estimation as well as the sequential operation of incremental clustering algorithm. Radial basis function neural networks consist of condition part, conclusion part and aggregation part. In the condition part, incremental K-means clustering algorithms is used tweights of the conclusion part are given as linear function and parameters are calculated using recursive least squareo get the center points of data and find the fitness using gaussian function as the activation function. Connection s estimation. In the aggregation part, a final output is obtained by center of gravity method. Using machine learning data, performance index are shown and compared with other models. Also, the performance of the incremental K-means clustering based-RBFNNs is carried out by using PSO. This study demonstrates that the proposed model shows the superiority of algorithmic design from the viewpoint of on-line processing for big data.

Unsupervised clustering for building a learning database of acoustic emission signals to identify damage mechanisms in unidirectional laminates

This work aims at developing a new clustering method of acoustic emission (AE) signals, called Incremental Clustering (IC), induced by damage mechanisms of glass fibre reinforced composite materials. The advantage of this method over other methods from literature is the capability to identify the signals carrying information and to provide the types of damage mechanisms without using additional expertise. To apply the method based on actual data, several specimens of glass fibre reinforced epoxy composites have been manufactured and subjected to specific mechanical tests. The developed method was compared to the k-means method that is extensively used to classify the AE signals. The reliability of the learning database was checked by the performance evaluation of the k Nearest Neighbours’ (kNN) classifiers. The kNN classifiers were trained by the training dataset and evaluated by the test dataset. The area under the receiver operating characteristic curves (AUC) was used as a criterion for evaluating the performance of classifiers. From specific mechanical tests, the IC method presented more advantages to successfully classify the AE signals and build a labelled learning database than the k-means method. The chronology of appearance of different damage mechanisms demonstrated the effectiveness of the incremental method. The powerful performance of the supervised method, characterized by values of AUC greater than 0.9 for each damage type, confirmed the reliability of the obtained learning database.

Online and Scalable Unsupervised Network Anomaly Detection Method

Nowadays, network intrusion detectors mainly rely on knowledge databases to detect suspicious traffic. These databases have to be continuously updated which requires important human resources and time. Unsupervised network anomaly detectors overcome this issue by using “intelligent” techniques to identify anomalies without any prior knowledge. However, these systems are often very complex as they need to explore the network traffic to identify flows patterns. Therefore, they are often unable to meet real-time requirements. In this paper, we present a new online and real-time unsupervised network anomaly detection algorithm (ORUNADA). Our solution relies on a discrete time-sliding window to update continuously the feature space and an incremental grid clustering to detect rapidly the anomalies. The evaluations showed that ORUNADA can process online large network traffic while ensuring a low detection delay and good detection performance. The experiments performed on the traffic of a core network of a Spanish intermediate Internet service provider demonstrated that ORUNADA detects in less than half a second an anomaly after its occurrence. Furthermore, the results highlight that our solution outperforms in terms of true positive rate and false positive rate existing techniques reported in the literature.

Large-scale fuzzy multiple-medoid clustering method

As one of feasible clustering techniques for large-scale data, incremental fuzzy clustering, which copes with data in chunks, has triggered more attentions in recent years. The existing methods, such as online fuzzy C-medoids (OFCMd) and history-based online fuzzy C-medoids (HOFCMd), employ only on e medoid to represent each cluster in chunks. Due to the fact that the representativeness of the one-medoid modality is sometimes unsatisfactory, a novel large-scale fuzzy multiple-medoid clustering (LS-FMMdC) method is presented to strengthen the clustering effectiveness for large-scale data. The performance of the proposed method is verified by comparing LS-FMMdC with OFCMd and HOFCMd on both synthetic and real-life large-scale data sets.

Coregistration of Satellite Images and Airborne LiDAR Data Through the Automatic Bias Reduction of RPCs

Rational polynomial coefficients (RPCs), which are provided for most of the commercial satellite images, match relevant image locations to their three-dimensional-ground positions. In spite of their key role in most of remote sensing analysis, a considerable amount of bias is usually seen in the RPCs that necessitate their refinement as a common preprocessing. This refinement is typically performed using some complimentary control information (e.g., ground control points). This paper proposes a matching scheme between the georeferenced airborne LiDAR data and high-resolution satellite images (HRSI) to automatically obtain the control data that is required for RPCs bias compensation. The main contribution of this paper is the design and implementation of a shadow-based matching strategy to correspond the inherently different HRSI and LiDAR data. In this process, RPCs are regarded as the initial relating model between HRSI and LiDAR data and the existing bias of these parameters are ultimately reduced. A novel method, called incremental clustering, is used to automatically detect the shadow cast from the HRSI. In addition, a geometrical method is developed for shadow reconstructions from LiDAR data. These generated shadow maps are then assimilated from geometrical point of view as well as coordinate systems. Finally, a frequency domain matching is performed to find and compensate for the existent bias in RPCs. The obtained results indicate a precise bias reduction of the RPCs, where the initial misalignment of georeferencing is improved from 18 m (30 pixels) to about 0.58 m (1 pixel).

On the Accuracy and Parallelism of GPGPU-Powered Incremental Clustering Algorithms.

Incremental clustering algorithms play a vital role in various applications such as massive data analysis and real-time data processing. Typical application scenarios of incremental clustering raise high demand on computing power of the hardware platform. Parallel computing is a common solution to meet this demand. Moreover, General Purpose Graphic Processing Unit (GPGPU) is a promising parallel computing device. Nevertheless, the incremental clustering algorithm is facing a dilemma between clustering accuracy and parallelism when they are powered by GPGPU. We formally analyzed the cause of this dilemma. First, we formalized concepts relevant to incremental clustering like evolving granularity. Second, we formally proved two theorems. The first theorem proves the relation between clustering accuracy and evolving granularity. Additionally, this theorem analyzes the upper and lower bounds of different-to-same mis-affiliation. Fewer occurrences of such mis-affiliation mean higher accuracy. The second theorem reveals the relation between parallelism and evolving granularity. Smaller work-depth means superior parallelism. Through the proofs, we conclude that accuracy of an incremental clustering algorithm is negatively related to evolving granularity while parallelism is positively related to the granularity. Thus the contradictory relations cause the dilemma. Finally, we validated the relations through a demo algorithm. Experiment results verified theoretical conclusions.

Representative Points and Cluster Attributes Based Incremental Sequence Clustering Algorithm

In order to improve the execution time and clustering quality of sequence clustering algorithm in large-scale dynamic dataset, a novel algorithm RPCAISC (Representative Points and Cluster Attributes Based Incremental Sequence Clustering) was presented. In this paper, density factor is defined. The primary representative point that has a density factor less than the prescribed threshold will be deleted directly. New representative points can be reselected from non-representative points. Moreover, the representative points of each cluster are modeled using the K-nearest neighbor method. The definition of the relevant degree (RD) between clusters was also proposed. The RD is computed by comprehensively considering the correlations of objects within a cluster and between different clusters. Then, whether the two clusters need to merge is determined. Additionally, the cluster attributes of the initial clustering are retained with this process. By calculating the matching degree between the incremental sequence and the existing cluster attributes, dynamic sequence clustering can be achieved. The theoretic experimental results and analysis prove that RPCAISC has better correct rate of clustering results and execution efficiency.

Incremental topological spatial association rule mining and clustering from geographical datasets using probabilistic approach

Due to the dynamic updating of real time spatial databases, the preservation of spatial association rules for dynamic database is a vital issue because the updates may not only invalidate some existing rules but also make other rules relevant. Consequently, the dynamic updating of spatial rules was handled by many researchers through the incremental association rule mining algorithm. Accordingly, in this paper we have developed an incremental topological association rule mining of geographical datasets using probabilistic approach. Initially, the spatial database is read out and it is passed through probability-based incremental association rule discovery algorithm to mine the topological spatial association rules. Once the rules are mined from the spatial database, the assumption here is that the database is dynamically updating for every time interval. In order to handle this dynamic nature, the proposed incremental topological association rule mining process is used in this paper. Here, the candidate topological rule generation is done from the spatial association rules using the topological relations such as, nearby, disjoint, intersects and inside/outside and the topological support is calculated using the proposed probabilistic topological support model. Finally, the spatial clustering is performed based on the mined spatial rules. From the experimentation, we proved that the maximum accuracy reached by the proposed method is 83.14% which is higher than the existing methods, which is defined as the ratio of the occurred rules and total number of topological data objects.

Information Clustering Using Manifold-Based Optimization of the Bag-of-Features Representation.

In this paper, a manifold-based dictionary learning method for the bag-of-features (BoF) representation optimized toward information clustering is proposed. First, the spectral representation, which unwraps the manifolds of the data and provides better clustering solutions, is formed. Then, a new dictionary is learned in order to make the histogram space, i.e., the space where the BoF historgrams exist, as similar as possible to the spectral space. The ability of the proposed method to improve the clustering solutions is demonstrated using a wide range of datasets: two image datasets, the 15-scene dataset and the Corel image dataset, one video dataset, the KTH dataset, and one text dataset, the RT-2k dataset. The proposed method improves both the internal and the external clustering criteria for two different clustering algorithms: 1) the -means and 2) the spectral clustering. Also, the optimized histogram space can be used to directly assign a new object to its cluster, instead of using the spectral space (which requires reapplying the spectral clustering algorithm or using incremental spectral clustering techniques). Finally, the learned representation is also evaluated using an information retrieval setup and it is demonstrated that improves the retrieval precision over the baseline BoF representation.

An intelligent system for mineral identification in thin sections based on a cascade approach

In this study, an intelligent system for mineral identification in thin sections is proposed based on RGB and HSI color spaces and texture features in plane and cross polarized light. The proposed system has two phases for mineral identification. In phase#1, which is the segmentation phase, 12 color components are extracted for each pixel, and using an incremental clustering algorithm, several mineral clusters including index mineral are produced. Afterwards, in phase#2 which is the identification phase, the produced mineral clusters are identified based on a cascade classification approach. The first level of the cascade includes a set of artificial neural networks (ANNs) corresponding to the number of input minerals which are trained based on color components. In the first level, those minerals exhibiting different colors in plane or cross polarized light are identified. The second level of the cascade includes one ANN which is trained based on texture features in plane and cross polarized light images. In the second level, those minerals which are indistinguishable based on color components in both plane and cross polarized light are identified (are rejected in the first level of the cascade). The final output of the system is the name and number of minerals, boundary and percentage of each mineral in thin section, and eventually the name of probable target rock. The proposed system is able to recognize 23 test igneous minerals with the overall accuracy of 93.81%. The proposed system can be applied in important applications which require a real time segmentation and identification map such as petrography, and NASA Mars Explorations.

Tune Up Fuzzy C-Means for Big Data: Some Novel Hybrid Clustering Algorithms Based on Initial Selection and Incremental Clustering

Data are getting larger, and most of them are necessary for our businesses. Rapid explosion of data brings us a number of challenges relating to its complexity and how the most important knowledge can be captured in reasonable time. Fuzzy C-means (FCM)—one of the most efficient clustering algorithms which have been widely used in pattern recognition, data compression, image segmentation, computer vision and many other fields—also faces the problem of processing large datasets. In this paper, we propose some novel hybrid clustering algorithms based on incremental clustering and initial selection to tune up FCM for the Big Data problem. The first algorithm determines meshes of rectangle covering data points as the representatives, while the second one considers data points that have high influence to others as the representatives. The representatives are then clustered by FCM, and the new centers are selected as initial ones for clustering of the dataset. Theoretical analyses of the new algorithms including comparison of quality of solutions when clustering the representatives set versus the entire set are examined. The experimental results on both simulated and real datasets show that total computational time of the new methods including time of finding representatives and clustering is faster than those of other relevant algorithms. The validation on clustering quality is also examined. The findings of this paper have great impact and significance to researches in the fields of soft computing and Big Data processing. It is obvious that computing methodologies nowadays are facing with huge amount of diverse and complex data structures. Speed of processing is the main priority when considering effectiveness of a specific method. The findings demonstrated practical algorithms and investigated their characteristics that could be referenced by other researchers in similar applications. The usefulness and significance of this research are clearly demonstrated within the extent of real-life applications.

Supervised Adaptive Incremental Clustering for data stream of chunks

Many supervised clustering algorithms have been developed to find the optimal clusters for static datasets by presetting some parameters, but they are seldom suitable for dynamic datasets, such as the data stream of chunks. To find the optimal clusters of the data stream of chunks, a novel Supervised Adaptive Incremental Clustering (SAIC) algorithm is proposed. SAIC can cluster dynamic datasets of arbitrary shapes and sizes automatically. It includes learning and post-processing phases. In the learning phase, each cluster updates adaptively according to its learning rate that is calculated from its counter value. All data points are shuffled at each iteration in order to make SAIC insensitive to the input order of data points. In the post-processing phase, the outliers or boundary points are eliminated according to the counter value of each cluster and the number of iterations. Four synthetic datasets and fourteen UCI datasets are used to evaluate the performance of SAIC, respectively. The experiments on UCI datasets show that SAIC reaches to or outperforms some other supervised clustering algorithms and several unsupervised incremental clustering algorithms. In addition, three data stream of chunks are used to evaluate SAIC from different aspects, which shows SAIC has the scalability and incremental learning ability for the clustering of data streams of chunks.

Incremental Clustering-Based Facial Feature Tracking Using Bayesian ART

Person-independent, emotion specific facial feature tracking have been of interest in the machine vision society for decades. Among various methods, the constrained local model (CLM) has shown significant results in person-independent feature tracking. In 63this paper, we propose an automatic, efficient, and robust method for emotion specific facial feature detection and tracking from image sequences. Considering a 17-point feature model on the frontal face region, the proposed tracking framework incorporates CLM with two incremental clustering algorithms to increase robustness and minimize tracking error during feature tracking. The Patch Clustering algorithm is applied to build an appearance model of face frames by organizing previously encountered similar patches into clusters while the shape Clustering algorithm is applied to build a structure model of face shapes by organizing previously encountered similar shapes into clusters followed by Bayesian adaptive resonance theory (ART). Both models are used to explore the similar features/shapes in the successive images. The clusters in each model are built and updated incrementally and online, controlled by amount of facial muscle movement. The overall performance of the proposed incremental clustering-based facial feature tracking (ICFFT) is evaluated using the FGnet database and the extended Cohn-Kanade (CK+) database. ICFFT demonstrates better results than baseline-method CLM and provides robust tracking as well as improved localization accuracy of emotion specific facial features tracking.

Incremental kernel fuzzy c-means with optimizing cluster center initialization and delivery

Purpose The large volume of big data makes it impractical for traditional clustering algorithms which are usually designed for entire data set. The purpose of this paper is to focus on incremental clustering which divides data into series of data chunks and only a small amount of data need to be clustered at each time. Few researches on incremental clustering algorithm address the problem of optimizing cluster center initialization for each data chunk and selecting multiple passing points for each cluster. Design/methodology/approach Through optimizing initial cluster centers, quality of clustering results is improved for each data chunk and then quality of final clustering results is enhanced. Moreover, through selecting multiple passing points, more accurate information is passed down to improve the final clustering results. The method has been proposed to solve those two problems and is applied in the proposed algorithm based on streaming kernel fuzzy c-means (stKFCM) algorithm. Findings Experimental results show that the proposed algorithm demonstrates more accuracy and better performance than streaming kernel stKFCM algorithm. Originality/value This paper addresses the problem of improving the performance of increment clustering through optimizing cluster center initialization and selecting multiple passing points. The paper analyzed the performance of the proposed scheme and proved its effectiveness.

Multi-object tracking via discriminative appearance modeling

Tracking multiple objects is important for automatic video content analysis and virtual reality. Recently, how to formulate data association optimization more effectively to overcome ambiguous detected responses and how to build more effective association affinity model have attracted more concerns. To address these issues, we propose a metric learning and multi-cue fusion based hierarchical multiple hypotheses tracking method (MHMHT), which conducts data association more robustly and incorporates more temporal context information. The association appearance similarity is calculated using the distances between feature vectors in each associated tracklet and the salient templates of each track hypothesis, which is then fused with the dynamic similarity calculated according to Kalman filter online to get association affinity. To make appearance similarity more discriminative, the spatial-temporal relationships of reliable tracklets in sliding temporal window are used as constraints to learn the discriminative appearance metric which measures the distance between feature vectors and salient templates. The salient templates of generated track hypotheses are updated using an incremental clustering method, considering the high order temporal context information. We evaluate our MHMHT tracker on challenging benchmark datasets. Qualitative and quantitative evaluations demonstrate that the proposed tracking algorithm performs favorably against several state-of-the-art methods.

A NOVEL INCREMENTAL CLUSTERING FOR INFORMATION EXTRACTION FROM SOCIAL NETWORKS

The challenge of this project concentrates upon the issue of synopsis on the remark string regarding the particular message from social media. Because of the more fame of social media, amount of remarks may increment by the side of more ratio directly later the societal message is printed. Clients can want to achieve the detailed comprehension about remark string without study entire remark set, an attempt is made in order to bunch remarks by comparative substance all at once also produce the succinct judgment outline only for this message. Seeing that any time various clients can ask for a synopsis outcome, but the existing clustering strategies cannot fulfill the current requirement of this program. We design an incremental bunching issue for remark string synopsis upon social media also propose Incremental Clustering method it can incrementally bring up to date bunching outcomes including recent arriving remarks. And also, we design a presentation interface comprising of fundamental data, keyterms, and delegate remarks. This brief look presentation interface assists clients to rapidly achieve the outline comprehension about remark string. From the experimental results it is observed that Incremental Clustering method is more efficient than KMeans and Batch clustering methods.

Grouping of business processes models based on an incremental clustering algorithm using fuzzy similarity and multimodal search

Nowadays, many companies standardize their operations through Business Process (BP), which are stored in repositories and reused when new functionalities are required. However, finding specific processes may become a cumbersome task due to the large size of these repositories. This paper presents MulTimodalGroup, a model for grouping and searching business processes. The grouping mechanism is built upon a clustering algorithm that uses a similarity function based on fuzzy logic; this grouping is performed using the results of each user request. By its part, the search is based on a multimodal representation that integrates textual and structural information of BP. The assessment of the proposed model was carried out in two phases: 1) internal quality assessment of groups and 2) external assessment of the created groups compared with an ideal set of groups. The assessment was performed using a closed BP collection designed collaboratively by 59 experts. The experimental results in each phase are promising and evidence the validity of the proposed model.

Improving hot region prediction by parameter optimization of density clustering in PPI

This paper proposed an optimized algorithm which combines density clustering of parameter selection with feature-based classification for hot region prediction. First, all the residues are classified by SVM to remove non-hot spot residues, then density clustering of parameter selection is used to find hot regions. In the density clustering, this paper studies how to select input parameters. There are two parameters radius and density in density-based incremental clustering. We firstly fix density and enumerate radius to find a pair of parameters which leads to maximum number of clusters, and then we fix radius and enumerate density to find another pair of parameters which leads to maximum number of clusters. Experiment results show that the proposed method using both two pairs of parameters provides better prediction performance than the other method, and compare these two predictive results, the result by fixing radius and enumerating density have slightly higher prediction accuracy than that by fixing density and enumerating radius.

A comparison study of clustering algorithms for microblog posts

Clustering is a popular unsupervised learning approach for topic analysis in text mining. In this paper, we do a comparison study of clustering algorithms for microblog posts, including weighting and programming model. Our experimental data is crawled from Sina Weibo in China. They are the 74,662 microblogs of 14 topics about Internet Technology. First of all, we do preprocessing to these microblog posts. Then we propose a manual sampling based dynamic incremental clustering algorithm (MS-DICA) to extract the topic threads from the microblogs we crawled. We evaluate the proposed algorithm from four aspects. Moreover, experimental comparisons are done in terms of accuracy and efficiency with the traditional k-means algorithm. Our experimental results show that the proposed MS-DICA is effective in the topic thread extraction. Its accuracy is close to the traditional k-means algorithm, and the running speed improves more than five times. In addition, the MapReduce programming model in Hadoop distributed computation platform that can run paralleled the k-means algorithm for cluster speeding up.

An incremental nonparametric Bayesian clustering-based traversable region detection method

Navigation capability in complex and unknown outdoor environments is one of the major requirements for an autonomous vehicle and a robot that perform tasks such as a military mission or planetary exploration. Robust traversability estimation in unknown environments would allow the vehicle or the robot to devise control and planning strategies to maximize their effectiveness. In this study, we present a self-supervised on-line learning architecture to estimate the traversability in complex and unknown outdoor environments. The proposed approach builds a model by clustering appearance data using the newly proposed incremental nonparametric Bayesian clustering algorithm. The clusters are then classified as being either traversable or non-traversable. Because our approach effectively groups unknown regions with similar properties, while the vehicle is in motion without human intervention, the vehicle can be deployed to new environments by automatically adapting to changing environmental conditions. We demonstrate the performance of the proposed clustering algorithm through intensive experiments using synthetic and real data and evaluate the viability of the traversability estimation using real data sets collected in outdoor environment.