Clusters In Dataset Research Articles

Nonparametric REML-like Estimation in Linear Mixed Models with Uncorrelated Homoscedastic Errors

Restricted Maximum Likelihood (REML) is the most recommended approach for fitting a Linear Mixed Model (LMM) nowadays. Yet, as ML, REML suffers the drawback that it performs such a fitting by assuming normality for both the random effects and the residual errors, a dubious assumption for many real data sets. Now, there have been several attempts at trying to justify the use of the REML likelihood equations outside of the Gaussian world, with varying degrees of success. Recently, a new fitting methodology, code named 3S, was presented for LMMs with only added assumption (to the basic ones) that the residual errors are uncorrelated and homoscedastic. Specifically, the 3S-A1 variant was designed and then shown, for Gaussian LMMs, to differ only slightly from ML estimation. In this article, using the same 3S framework, we develop another iterative nonparametric estimation methodology, code named 3S-A1.RE, for the kind of LMMs just mentioned. However, we show that if the LMM is, indeed, Gaussian with i.i.d. residual errors, then the set of estimating equations defining any 3S-A1.RE iterative procedure is equivalent to the set of REML equations, but while including the nonnegativity constraints on all variance estimates, as well as positive semi-definiteness on all covariance matrices. In numerical tests on some simulated and real world clustered and longitudinal data sets, our new methods proved to be highly competitive when compared to the traditional REML in the R statistical software.

Large-Scale Fuzzy Least Squares Twin SVMs for Class Imbalance Learning

Twin support vector machines (TSVMs) have been successfully employed for binary classification problems. With the advent of machine learning algorithms, data have proliferated and there is a need to handle or process large-scale data. TSVMs are not successful in handling large-scale data due to the following: 1) the optimization problem solved in the TSVM needs to calculate large matrix inverses, which makes it an ineffective choice for large-scale problems; 2) the empirical risk minimization principle is employed in the TSVM and, hence, may suffer due to overfitting; and 3) the Wolfe dual of TSVM formulation involves positive-semidefinite matrices, and hence, singularity issues need to be resolved manually. Keeping in view the aforementioned shortcomings, in this article, we propose a novel large-scale fuzzy least squares TSVM for class imbalance learning (LS-FLSTSVM-CIL). We formulate the LS-FLSTSVM-CIL such that the proposed optimization problem ensures that: 1) no matrix inversion is involved in the proposed LS-FLSTSVM-CIL formulation, which makes it an efficient choice for large-scale problems; 2) the structural risk minimization principle is implemented, which avoids the issues of overfitting and results in better performance; and 3) the Wolfe dual formulation of the proposed LS-FLSTSVM-CIL model involves positive-definite matrices. In addition, to resolve the issues of class imbalance, we assign fuzzy weights in the proposed LS-FLSTSVM-CIL to avoid bias in dominating the samples of class imbalance problems. To make it more feasible for large-scale problems, we use an iterative procedure known as the sequential minimization principle to solve the objective function of the proposed LS-FLSTSVM-CIL model. From the experimental results, one can see that the proposed LS-FLSTSVM-CIL demonstrates superior performance in comparison to baseline classifiers. To demonstrate the feasibility of the proposed LS-FLSTSVM-CIL on large-scale classification problems, we evaluate the classification models on the large-scale normally distributed clustered (NDC) dataset. To demonstrate the practical applications of the proposed LS-FLSTSVM-CIL model, we evaluate it for the diagnosis of Alzheimer’s disease and breast cancer disease. Evaluation on NDC datasets shows that the proposed LS-FLSTSVM-CIL has feasibility in large-scale problems as it is fast in comparison to the baseline classifiers.

An Effective Secured Dynamic Network-Aware Multi-Objective Cuckoo Search Optimization for Live VM Migration in Sustainable Data Centers

With the increasing use of cloud computing by organizations, cloud data centers are proliferating to meet customers’ demands and host various applications using virtual machines installed in physical servers. Through Live Virtual Machine Migration (LVMM) methods, cloud service providers can provide improved computing capabilities for server consolidation maintenance of systems and potential power savings through a reduction in the distribution process to customers. However, Live Virtual Machine Migration has its challenges when choosing the best network path for maximizing the efficiency of resources, reducing consumption, and providing security. Most research has focused on the load balancing of resources and the reduction in energy consumption; however, they could not provide secure and optimal resource utilization. A framework has been created for sustainable data centers that pick the most secure and optimal dynamic network path using an intelligent metaheuristic algorithm, namely, the Network-aware Dynamic multi-objective Cuckoo Search algorithm (NDCS). The developed hybrid movement strategy enhances the search capability by expanding the search space and adopting a combined risk score estimation of each physical machine (PM) as a fitness criterion for ensuring security with rapid convergence compared to the existing strategies. The proposed method was assessed using the Google cluster dataset to ascertain its worthiness. The experimental results show the supremacy of the proposed method over existing methods by ensuring services with a lower total migration time, lower energy consumption, less makespan time, and secure optimum resource utilization.

EHHR: an efficient evolutionary hyper-heuristic based recommender framework for short-text classifier selection.

With various machine learning heuristics, it becomes difficult to choose an appropriate heuristic to classify short-text emerging from various social media sources in the form of tweets and reviews. The No Free Lunch theorem asserts that no heuristic applies to all problems indiscriminately. Regardless of their success, the available classifier recommendation algorithms only deal with numeric data. To cater to these limitations, an umbrella classifier recommender must determine the best heuristic for short-text data. This paper presents an efficient reminisce-enabled classifier recommender framework to recommend a heuristic for new short-text data classification. The proposed framework, “Efficient Evolutionary Hyper-heuristic based Recommender Framework for Short-text Classifier Selection (EHHR),” reuses the previous solutions to predict the performance of various heuristics for an unseen problem. The Hybrid Adaptive Genetic Algorithm (HAGA) in EHHR facilitates dataset-level feature optimization and performance prediction. HAGA reveals that the influential features for recommending the best short-text heuristic are the average entropy, mean length of the word string, adjective variation, verb variation II, and average hard examples. The experimental results show that HAGA is 80% more accurate when compared to the standard Genetic Algorithm (GA). Additionally, EHHR clusters datasets and rank heuristics cluster-wise. EHHR clusters 9 out of 10 problems correctly.

Optimal k-means clustering using artificial bee colony algorithm with variable food sources length

<span>Clustering is a robust machine learning task that involves dividing data points into a set of groups with similar traits. One of the widely used methods in this regard is the k-means clustering algorithm due to its simplicity and effectiveness. However, this algorithm suffers from the problem of predicting the number and coordinates of the initial clustering centers. In this paper, a method based on the first artificial bee colony algorithm with variable-length individuals is proposed to overcome the limitations of the k-means algorithm. Therefore, the proposed technique will automatically predict the clusters number (the value of k) and determine the most suitable coordinates for the initial centers of clustering instead of manually presetting them. The results were encouraging compared with the traditional k-means algorithm on three real-life clustering datasets. The proposed algorithm outperforms the traditional k-means algorithm for all tested real-life datasets.</span>

An Effective Initialization Method Based on Quartiles for the K-means Algorithm

Objectives: This study aims to speed up the K-means algorithm by offering a deterministic quartile-based seeding strategy for initializing preliminary cluster centers for the K-means algorithm, enabling it to efficiently build high-quality clusters. Methods: We have investigated various cluster center initialization approaches in literature and presented our findings. For the Kmeans algorithm, we here propose a novel deterministic technique based on quartiles for finding initial cluster centers. To obtain the preliminary cluster centers, we have applied our suggested approach to the data set. The initial cluster centers determined by our suggested method are then entered into the K-means algorithm. The proposed seeding method is evaluated on sixteen benchmark clustering data sets: five synthetic and eleven real data sets. Python is used to run the simulation. Findings: Based on empirical results from experiments, it is evident that our proposed cluster center initialization method allows the K-means algorithm to form clusters with SSE values comparable to the minimum SSE values produced by repeated Random or Kmeans++ initializations. Furthermore, our deterministic initialization strategy assures that the K-means algorithm converges faster than the Random and K-means++ initialization techniques. Novelty: In this study, we explore the potential of quartile-based seeding as a technique of accelerating the Kmeans algorithm. Needless to add, as our seeding method is deterministic, the requirement to run K-means repeatedly with different stochastic initializations is completely eliminated. Also, our initialization strategy assures that there is remarkable saving in execution time as compared to the Random and Kmeans++ initialization techniques. Moreover, it is found that after initializing with our offered method, the solution obtained with just a single run of K-means produces optimal clusters. Applications: Our proposed seeding technique will be helpful for initializing the K-means algorithm in time-sensitive applications, applications managing large amounts of data, and applications looking for deterministic cluster solutions. Keywords: Kmeans Algorithm; Initialization Method; Speeding Kmeans; Quartiles; Clustering; Deterministic Initialization Method

Development zones and firms’ performance: the impact of development zones on firms’ performance for a Chinese industrial cluster

ABSTRACT This paper examines the policy effect of development zones on firms’ performance, excluding the agglomeration effect. We constructed an industrial cluster dataset on China’s manufacturing industries to identify the impacts on firms’ productivity and other performance indices after the establishment of development zones in 2006. Based on the estimated results obtained from a difference-in-differences analysis, development zones are conducive to promoting firms’ performance, and the policy effect is heterogeneous across industrial clusters, regions and firms. The findings drawn from this study can be beneficial to policymakers in their pursuit of promoting regional development through favourable industrial policies.

Geographical clusters of dengue outbreak in Singapore during the Covid-19 nationwide lockdown of 2020

Dengue, a mosquito-transmitted viral disease, has posed a public health challenge to Singaporean residents over the years. In 2020, Singapore experienced an unprecedented dengue outbreak. We collected a dataset of geographical dengue clusters reported by the National Environment Agency (NEA) from 15 February to 9 July in 2020, covering the nationwide lockdown associated with Covid-19 during the period from 7 April to 1 June. NEA regularly updates the dengue clusters during which an infected person may be tagged to one cluster based on the most probable infection location (residential apartment or workplace address), which is further matched to fine-grained spatial units with an average coverage of about 1.35 km2. Such dengue cluster dataset helps not only reveal the dengue transmission patterns, but also reflect the effects of lockdown on dengue spreading dynamics. The resulting data records are released in simple formats for easy access to facilitate studies on dengue epidemics.

Local Learning-based Multi-task Clustering

Clustering plays an essential role in machine learning and data mining. Many real-world datasets for clustering are often different but related in the big data era. Recent research suggests that the individual performance of each learning task could be significantly improved by appropriately transferring knowledge among the related tasks. However, traditional clustering methods, which is limited to a single task, often ignore the correlation between multiple clustering tasks. Multi-task clustering (MTC) has attracted widespread interest recently by attempting to mine sufficient knowledge within multiple related tasks. Existing MTC methods still have the following limitations: 1) the intrinsic geometry of data is seldom considered; 2) the discriminative low-dimensional representation of data is not well explored; 3) the cluster structure of data is neglected in the process of clustering. In order to tackle the above issues, we propose a novel end-to-end Local Learning-based Multi-task Clustering (LLMC) method, which can simultaneously explore discriminative information in a low-dimensional subspace and expose the clustering results for multiple tasks. In particular, the proposed LLMC method can effectively integrate transfer learning, subspace learning, local manifold learning, and clustering. Specifically, a joint projection of heterogeneous features is introduced to control the number of features shared by all the tasks to transfer knowledge among tasks. An efficient iterative algorithm is designed to optimize the objective and is theoretically guaranteed to be convergent. Experimental results demonstrate the superiority of the proposed method over the state-of-the-art single-task and multi-task clustering methods.

A deep matrix factorization based approach for single-cell RNA-seq data clustering.

The rapid development of single-cell sequencing technologies makes it possible to analyze cellular heterogeneity at the single-cell level. Cell clustering is one of the most fundamental and common steps in the heterogeneity analysis. However, due to the high noise level, high dimensionality and high sparsity, accurate cell clustering is still challengeable. Here, we present DeepCI, a new clustering approach for scRNA-seq data. Using two autoencoders to obtain cell embedding and gene embedding, DeepCI can simultaneously learn cell low-dimensional representation and clustering. In addition, the recovered gene expression matrix can be obtained by the matrix multiplication of cell and gene embedding. To evaluate the performance of DeepCI, we performed it on several real scRNA-seq datasets for clustering and visualization analysis. The experimental results show that DeepCI obtains the overall better performance than several popular single cell analysis methods. We also evaluated the imputation performance of DeepCI by a dedicated experiment. The corresponding results show that the imputed gene expression of known specific marker genes can greatly improve the accuracy of cell type classification.

Constraining spatial curvature with large-scale structure

We analyse the clustering of matter on large scales in an extension of the concordance model that allows for spatial curvature. We develop a consistent approach to curvature and wide-angle effects on the galaxy 2-point correlation function in redshift space. In particular we derive the Alcock-Paczynski distortion of fσ 8, which differs significantly from empirical models in the literature. A key innovation is the use of the 'Clustering Ratio', which probes clustering in a different way to redshift-space distortions, so that their combination delivers more powerful cosmological constraints. We use this combination to constrain cosmological parameters, without CMB information. In a curved Universe, we find that Ωm, 0=0.26± 0.04 (68% CL). When the clustering probes are combined with low-redshift background probes — BAO and SNIa — we obtain a CMB-independent constraint on curvature: Ω K, 0 = 0.0041-0.0504 +0.0500. We find no Bayesian evidence that the flat concordance model can be rejected. In addition we show that the sound horizon at decoupling is r d = 144.57 ± 2.34 Mpc, in agreement with its measurement from CMB anisotropies. As a consequence, the late-time Universe is compatible with flat ΛCDM and a standard sound horizon, leading to a small value of H 0, without assuming any CMB information. Clustering Ratio measurements produce the only low-redshift clustering data set that is not in disagreement with the CMB, and combining the two data sets we obtain Ω K, 0 = -0.023 ± 0.010.

The Initialization of Flexible K-Medoids Partitioning Method Using a Combination of Deviation and Sum of Variable Values

This research proposed a new algorithm for clustering datasets using the Flexible K-Medoids Partitioning Method. The procedure is divided into two phases, selecting the initial medoids and determining the partitioned dataset. The initial medoids are selected based on the block representation of a combination of the sum and deviation of the variable values. The relative positions of the objects will be separated when the sum of the values of the p variables is different even though these objects have the same variance. The objects are selected flexibly from each block as the initial medoids to construct the initial groups. This process ensures that any identical objects will be in the same group. The candidate of final medoids is determined randomly by selecting objects from each initial group. Then, the final medoids were identified based on the combination of objects that produces the minimum value of the total deviation within the cluster. The proposed method overcomes the empty group that may arise in a simple and fast k-medoids algorithm. In addition, it overcomes identical objects in the different groups that may occur in the initialization of the simple k-medoids algorithm. Furthermore, the artificial data and six real datasets, namely iris, ionosphere, soybean small, primary tumor, heart disease case 1 and zoo were used to evaluate this method, and the results were compared with other algorithms based on the initial and final groups' performance. The experiment results showed that the proposed method ensures that no initial groups are empty. For real datasets, the adjusted Rand index and clustering accuracy of the final groups of the new algorithm outperforms the other methods.

DDDQN‐TS: A task scheduling and load balancing method based on optimized deep reinforcement learning in heterogeneous computing environment

Task scheduling and load balancing problem of heterogeneous computing environment (HCE) is getting more and more attention these days and has become a research hotspot in this field. The task scheduling and load balancing problem of heterogeneous environment, which refers to assigning a set of tasks to a specific set of machines with different hardware and different computing performance with the goal of minimizing task processing time and keeping load balance among machines, has been proved to be an NP-complete problem. The development of artificial intelligence provides new ideas to solve this problem. In this paper, we propose a novel task scheduling and load balancing method based on optimized deep reinforcement learning in HCE. First, we formulate task scheduling problem as a Markov decision process and then adopt a dueling double deep Q-learning network to search the optimal task allocation solution. Then we use two well-known large-scale cluster data sets Google Cloud Jobs data set and Alibaba Cluster Trace data set to validate our approach. The experimental results show that compared with other existing solutions, our proposed method can achieve much shorter task response time and better load balancing effect.

Enhanced LSTM-based community energy consumption prediction model leveraging shared building cluster datasets

Unlike in the past, as the scale of buildings expands from a single building to a group of buildings, it is necessary to respond to the demand for energy consumption for the entire community. This study attempted to improve the electricity consumption prediction performance of single buildings by reflecting the inherent occupant behaviour patterns embedded in the shared electricity consumption data of single buildings. The proposed method utilized electricity consumption data for all other 7 target buildings as a part of LSTM model inputs in combination with meteorological data to predict the electricity consumption profile of a specific building. As a result, the proposed method reduced the error by utilizing the similar energy profiles of the surrounding buildings, *and the prediction performance of the total electricity consumption of the community was increased by 5.11% based on RMSE. The newly proposed method reflects the inherent occupant behaviour patterns and temporal electricity consumption patterns from energy profiles of nearby buildings.

LLAM-MDCNet for Detecting Remote Sensing Images of Dead Tree Clusters

Clusters of dead trees are forest fires-prone. To maintain ecological balance and realize its protection, timely detection of dead trees in forest remote sensing images using existing computer vision methods is of great significance. Remote sensing images captured by Unmanned aerial vehicles (UAVs) typically have several issues, e.g., mixed distribution of adjacent but different tree classes, interference of redundant information, and high differences in scales of dead tree clusters, making the detection of dead tree clusters much more challenging. Therefore, based on the Multipath dense composite network (MDCN), an object detection method called LLAM-MDCNet is proposed in this paper. First, a feature extraction network called Multipath dense composite network is designed. The network’s multipath structure can substantially increase the extraction of underlying and semantic features to enhance its extraction capability for rich-information regions. Following that, in the row, column, and diagonal directions, the Longitude Latitude Attention Mechanism (LLAM) is presented and incorporated into the feature extraction network. The multi-directional LLAM facilitates the suppression of irrelevant and redundant information and improves the representation of high-level semantic feature information. Lastly, an AugFPN is employed for down-sampling, yielding a more comprehensive representation of image features with the combination of low-level texture features and high-level semantic information. Consequently, the network’s detection effect for dead tree cluster targets with high-scale differences is improved. Furthermore, we make the collected high-quality aerial dead tree cluster dataset containing 19,517 images shot by drones publicly available for other researchers to improve the work in this paper. Our proposed method achieved 87.25% mAP with an FPS of 66 on our dataset, demonstrating the effectiveness of the LLAM-MDCNet for detecting dead tree cluster targets in forest remote sensing images.

Datasets for Data-Centric Classification and Clustering

Datasets for Data-Centric Classification and Clustering

Datasets for Data-Centric Classification and Clustering

Datasets for Data-Centric Classification and Clustering

Datasets for Data-Centric Classification and Clustering

Datasets for Data-Centric Classification and Clustering

Serial crystallography with multi-stage merging of thousands of images.

KAMO and BLEND provide particularly effective tools to automatically manage the merging of large numbers of data sets from serial crystallography. The requirement for manual intervention in the process can be reduced by extending BLEND to support additional clustering options such as the use of more accurate cell distance metrics and the use of reflection-intensity correlation coefficients to infer `distances' among sets of reflections. This increases the sensitivity to differences in unit-cell parameters and allows clustering to assemble nearly complete data sets on the basis of intensity or amplitude differences. If the data sets are already sufficiently complete to permit it, one applies KAMO once and clusters the data using intensities only. When starting from incomplete data sets, one applies KAMO twice, first using unit-cell parameters. In this step, either the simple cell vector distance of the original BLEND or the more sensitive NCDist is used. This step tends to find clusters of sufficient size such that, when merged, each cluster is sufficiently complete to allow reflection intensities or amplitudes to be compared. One then uses KAMO again using the correlation between reflections with a common hkl to merge clusters in a way that is sensitive to structural differences that may not have perturbed the unit-cell parameters sufficiently to make meaningful clusters. Many groups have developed effective clustering algorithms that use a measurable physical parameter from each diffraction still or wedge to cluster the data into categories which then can be merged, one hopes, to yield the electron density from a single protein form. Since these physical parameters are often largely independent of one another, it should be possible to greatly improve the efficacy of data-clustering software by using a multi-stage partitioning strategy. Here, one possible approach to multi-stage data clustering is demonstrated. The strategy is to use unit-cell clustering until the merged data are sufficiently complete and then to use intensity-based clustering. Using this strategy, it is demonstrated that it is possible to accurately cluster data sets from crystals that have subtle differences.

MMAN: Metapath Based Multi-Level Graph Attention Networks for Heterogeneous Network Embedding (Student Abstract)

Current Heterogeneous Network Embedding (HNE) models can be roughly divided into two types, i.e., relation-aware and metapath-aware models. However, they either fail to represent the non-pairwise relations in heterogeneous graph, or only capable of capturing local information around target node. In this paper, we propose a metapath based multilevel graph attention networks (MMAN) to jointly learn node embeddings on two substructures, i.e., metapath based graphs and hypergraphs extracted from original heterogeneous graph. Extensive experiments on three benchmark datasets for node classification and node clustering demonstrate the superiority of MMAN over the state-of-the-art works.