High-quality Clusters Research Articles

Efficient Parallel Processing of k-Nearest Neighbor Queries by Using a Centroid-based and Hierarchical Clustering Algorithm

The k-Nearest Neighbor method is one of the most popular techniques for both classification and regression purposes. Because of its operation, the application of this classification may be limited to problems with a certain number of instances, particularly, when run time is a consideration. However, the classification of large amounts of data has become a fundamental task in many real-world applications. It is logical to scale the k-Nearest Neighbor method to large scale datasets. This paper proposes a new k-Nearest Neighbor classification method (KNN-CCL) which uses a parallel centroid-based and hierarchical clustering algorithm to separate the sample of training dataset into multiple parts. The introduced clustering algorithm uses four stages of successive refinements and generates high quality clusters. The k-Nearest Neighbor approach subsequently makes use of them to predict the test datasets. Finally, sets of experiments are conducted on the UCI datasets. The experimental results confirm that the proposed k-Nearest Neighbor classification method performs well with regard to classification accuracy and performance.

A Radio Frequency Energy Harvesting-Based Multihop Clustering Routing Protocol for Cognitive Radio Sensor Networks

In radio frequency energy harvesting-cognitive radio sensor networks (RF EH-CRSNs), clustering routing protocols can help deliver monitored data towards the sink and they are vital for network performance. However, due to the impact of dynamic channel availability, limited node transmission range, and position-dependent energy arrival, existing clustering routing protocols for nonEH-CRSNs and EH nonCRSNs cannot be applied to RF EH-CRSNs. In order to solve above challenges, an RF EH-based multihop clustering routing protocol (RFMCRP) based on nonlinear EH model is proposed in this paper. Firstly, by leveraging curve fitting tool and statistical analysis, the most reasonable nonlinear EH model is identified and it is utilized by RFMCRP to measure the harvested energy accurately. Secondly, the optimal number of clusters is theoretically derived and its value is used as benchmark to evaluate the validity of our proposal. Thirdly, energy control mechanism is introduced to manage node state, and it can help improve the stability of cluster construction. In addition, energy level function-based selection criteria are defined to select high-quality cluster heads and relays, which can help improve the energy sustainability and connectivity of the whole network. Simulation results show that RFMCRP gains obvious advantages over existing protocols in extending network lifetime and improving network monitoring capability.

Efficient and Privacy Preserving Clustering Algorithm for Spatiotemporal Data

The efficiency of a spatiotemporal data analysis algorithm decreases as the amount of data increases. Many clustering techniques have been proposed for data analysis applications. However, applying those techniques to spatiotemporal data clustering is still in its infancy. In this paper, we tackle the issue of clustering spatiotemporal data on public Cloud based on the distance between them. To increase the efficiency of spatiotemporal clustering, we have proposed a MapReduce-based framework for clustering. However, as spatiotemporal dataset contains sensitive information, directly outsourcing spatiotemporal data to Cloud servers will raise privacy concerns. To address the problem of privacy, we have proposed a privacy preserving clustering algorithm based on MapReduce for spatiotemporal data that can be efficiently outsourced for data processing on the Cloud servers. The proposed scheme allows the clustering operation to be performed directly on the encrypted spatiotemporal data by Cloud server. Extensive experimental evaluation with trajectory data shows that our scheme efficiently produces higher quality clustering results.

A ranking paired based artificial bee colony algorithm for data clustering

Data clustering aims to partition a dataset into k subsets according to a prespecified similarity measure. It is NP-hard, and has lots of real applications. This paper presents a ranking paired based artificial bee colony algorithm (RPABC) to solve data clustering. First, a chaotic map is employed to generate initial food sources. Second, in order to speed up the search, RPABC uses a ranking paired learning strategy to produce new positions. Finally, the best food source is utilised to guide the search in the onlooker bees' phase. Several datasets from the literature are used to test the RPABC. The computational results show that the proposed method is able to provide high quality clusters, and is more stable than the compared algorithms.

NLP Based Retrieval of Semantically Similar Private Tax Rulings

We report preliminary findings of the use of Natural Language Processing (NLP) methods for retrieving semantically similar private tax rulings (PTRs). The PTR corpus contains now more than 400 thousands of PTRs and was first described quantitively in [17]. Experiments show that use of BERT-based embeddings and cosine similarity results in very high quality set of legally similar PTRs. The quality of the set is at least on par with a manual search performed by a certified tax advisor. We also show that similar PTRs form high quality clusters that can be used for investigation of fine structure of the set. The results presented in the paper, although preliminary, already have significant practical value for tax practitioners.

Hierarchical Trust-Tech-Enhanced K-Means Methods and Their Applications to Power Grids

K-means has been widely used in solving a wide range of clustering problems arising in engineering and industrial applications, but it still suffers from several issues. To address these issues, a hierarchical K-means method enhanced by Trust-Tech (H-KTT) is presented in this paper. The proposed H-KTT method is composed of two stages. The first stage of H-KTT is a hierarchical K-means (H-K-means) method for enhancing K-means with better initial points. Second, the H-K-means method is further enhanced to find multiple high-quality clustering results by the Trust-Tech methodology. The H-KTT method was evaluated on several test datasets including the clustering of Automatic Meter Reading (AMR), popular in power grids, with promising results. In particular, the evaluation results indicate that the proposed H-KTT method can significantly improve both the quality and stability of the clustering results by the K-means method. Furthermore, while the K-means gives stochastic clustering results, the proposed H-KTT method usually gives deterministic clustering results.

Firefly Algorithm with Mini Batch K-Means Entropy Measure for Clustering Heterogeneous Categorical Timber Data

Clustering analysis is the process of identifying similar patterns in various types of data. Heterogeneous categorical data consists of data on ordinal, nominal, binary, and Likert scales. The clustering solution for heterogeneous data clustering remains difficult due to partitioning complex and dissimilarity features. It is necessary to find a solution to high-quality clustering techniques to efficiently determine the significant features of the data. This paper emphasizes using the firefly algorithm to reduce the distance gap between features and improve clustering performance. To obtain an optimal global solution for clustering, we proposed a hybrid of mini-batch k-means (MBK) clustering-based entropy distance measures (EM) with a firefly optimization algorithm (FA). This study compares the performance of hybrid K-Means, Agglomerative, DBSCAN, and Affinity clustering models with EM and FA. The evaluation uses a variety of data from the timber perception survey dataset. In terms of performance, the proposed MBK+EM+FA has superior and most effective clustering. It achieves a higher accuracy of 96.3 percent, a 97 percent F-measure, a 98 percent precision, and a 97 percent recall. Other external assessments revealed that the Homogeneity (HOMO) is 79.14 percent, the Fowlkes-Mallows Index (FMI) is 93.07 percent, the Completeness (COMP) is 78.04 percent, and the V-Measure (VM) is 78.58 percent. Both proposed MBK+EM+FA and MBK+EM took about 0.45s and 0.35s to compute, respectively. The excellent quality of the clustering results does not justify such time constraints. Surprisingly, the proposed model reduced the distance measure of all heterogeneous features. The future model could put heterogeneous categorical data from a different domain to the test.

Comparative Analysis of K-Means and Enhanced K-Means Algorithms for Clustering

Clustering in data mining is a way of organizing a set of objects in such a way that the objects in same bunch are more comparable and relevant to each other than to those objects in other bunches. In the modern information retrieval system, clustering algorithms are better if they result high quality clusters in efficient time. This study includes analysis of clustering algorithms k-means and enhanced k-means algorithm over the wholesale customers and wine data sets respectively. In this research, the enhanced k-means algorithm is found to be 5% faster for wholesale customers dataset for 4 clusters and 49%, 38% faster when the clusters size is increased to 8 and 13 respectively. The wholesale customers dataset when classified with 18 clusters the speedup was seen to be 29%. Similarly, in the case of wine dataset, the speed up is seen to be 10%, 30%, 49%, and 41% for 3, 8, 13 and 18 clusters respectively. Both of the algorithms are found very similar in terms of the clustering accuracy.

Electricity consumption pattern analysis beyond traditional clustering methods: A novel self-adapting semi-supervised clustering method and application case study

The fast-paced informatization of power systems across the world provides an unprecedented amount of data, which greatly facilitates their study and offers in turn the possibility to assist in the transition towards truly smart, low-carbon energy systems. In this context, the use of clustering methods for the study of household Electricity Consumption Behaviour (ECB) proves highly beneficial as it facilitates, among other things, more effective deployment of distributed renewable energy assets, development of differentiated tariff policies and load forecasting. However, the similarity metrics used in traditional clustering methods have difficulties in accurately capturing the time variability of electrical load profiles. In order to address this problem, we developed a novel semi-supervised automatic clustering method based on a self-adapting metric learning process. The proposed method is a bespoke application to the analysis of electricity demand load patterns that combines the recently developed Deep Linear Discriminant Analysis algorithm for supervised learning with the data-adaptive Affinity Propagation clustering algorithm (DLDA + AP), and achieves high-quality automatic clustering with an accuracy that is 75 percentage points higher than traditional methods such as k-means, on average. Based on this bespoke method, a unified load dictionary which captures the mainstream daily electricity consumption patterns of 5566 households in London was produced. Through the analysis of the load dictionary and household daily electricity consumption, it’s possible to build a complete ECB profile for the households in the sample dataset. Furthermore, combining the 206 household properties which were found to be strongly correlated with the ECB, this method provides a practical approach to residential customer segmentation for the electricity market.

Using a WeChat mini-program-based lactation consultant intervention to increase the consumption of mother’s own milk by preterm infants in the neonatal intensive care unit: a study protocol for a cluster randomized controlled trial

BackgroundThe benefits of mother’s own milk (MOM) for preterm infants have been widely recognized. Many studies have shown that the rate of breastfeeding of premature infants remains very low. Although many studies use measures to promote breastfeeding, few high-quality cluster randomized controlled studies have evaluated the effectiveness of these measures. WeChat is an instant messaging software for smart terminals, and WeChat mini-programs have been widely used to promote health and self-management in China. Based on this background, we designed a randomized controlled study based on WeChat mini-programs to promote MOM feeding of premature infants in the neonatal intensive care unit (NICU).Methods/designThis study will evaluate the effectiveness of WeChat mini-programs to increase the consumption of MOM feeding in twelve NICUs in Jiangsu Province, namely, six “intervention” NICUs and six “control” NICUs. The study process is as follows: (1) design and preparation, (2) NICU recruitment and training, (3) interpretation and analysis of baseline data, (4) quality control implementation process, and (5) data analysis feedback and publication of study reports. The primary outcome is the proportion of MOM feeding of premature infants during NICU hospitalization. The secondary outcomes are as follows: (1) time to initiation of MOM feeding (hours) and proportion of first-time MOM feeding (%), (2) duration of parenteral nutrition (days), (3) time to total gastrointestinal feeding (days), (4) hospitalization time and hospitalization cost, and (5) incidence of complications (necrotizing enterocolitis, bronchopulmonary dysplasia, feeding intolerance, late-onset sepsis, retinopathy of prematurity).DiscussionThis study is the first cluster randomized controlled trial on the intervention of using a WeChat mini-program-based lactation consultant for premature infants in the NICU in China. We hope this study can improve the consumption of MOM by NICU premature infants during hospitalization through the intervention of WeChat mini-programs.Trial registrationClinicalTrials.gov NCT04383379. Registered on May 5, 2020.

Multiway p-spectral graph cuts on Grassmann manifolds

Nonlinear reformulations of the spectral clustering method have gained a lot of recent attention due to their increased numerical benefits and their solid mathematical background. We present a novel direct multiway spectral clustering algorithm in the p-norm, for pin (1,2]. The problem of computing multiple eigenvectors of the graph p-Laplacian, a nonlinear generalization of the standard graph Laplacian, is recasted as an unconstrained minimization problem on a Grassmann manifold. The value of p is reduced in a pseudocontinuous manner, promoting sparser solution vectors that correspond to optimal graph cuts as p approaches one. Monitoring the monotonic decrease of the balanced graph cuts guarantees that we obtain the best available solution from the p-levels considered. We demonstrate the effectiveness and accuracy of our algorithm in various artificial test-cases. Our numerical examples and comparative results with various state-of-the-art clustering methods indicate that the proposed method obtains high quality clusters both in terms of balanced graph cut metrics and in terms of the accuracy of the labelling assignment. Furthermore, we conduct studies for the classification of facial images and handwritten characters to demonstrate the applicability in real-world datasets.

Construct high-precise SERS sensor by hierarchical superhydrophobic Si/Cu(OH)2 platform for ultratrace detection of food contaminants

Construction of the high-quality noble metal nanoparticle cluster (NMNPC) with big size for surface-enhanced Raman spectrum is still a challenge nowadays, due to its difficulty in integration of the uniformity, repeatability and sensibility. To realize this goal, a superhydrophobic platform composed of hierarchical Si/Cu(OH)2 nanowires with stable Cassie state was designed and prepared in this work. The optimal Ag NPC region is larger than 0.6 mm × 0.6 mm, and the limit of detection and relative standard deviation of it are respectively 2 × 10−18 M and 13.85%, indicating the fine SERS performance of it. Finite difference time domain simulations and systematic experimental analysis were made in detail here to figure out the underlying mechanism behind these phenomena. Furthermore, as a proof of concept, the Ag NPC was proved as an outstanding food contaminant sensor and successfully realized an ultratrace detection of both melamine in milk and malachite green on fish surface (ca 10−15 M, approaching to single-molecule level). This work provides a new way in constructing high-performance SERS substrate and shows promising application prospect in food safety inspection.

Fiber Clustering Acceleration With a Modified Kmeans++ Algorithm Using Data Parallelism

Fiber clustering methods are typically used in brain research to study the organization of white matter bundles from large diffusion MRI tractography datasets. These methods enable exploratory bundle inspection using visualization and other methods that require identifying brain white matter structures in individuals or a population. Some applications, such as real-time visualization and inter-subject clustering, need fast and high-quality intra-subject clustering algorithms. This work proposes a parallel algorithm using a General Purpose Graphics Processing Unit (GPGPU) for fiber clustering based on the FFClust algorithm. The proposed GPGPU implementation exploits data parallelism using both multicore and GPU fine-grained parallelism present in commodity architectures, including current laptops and desktop computers. Our approach implements all FFClust steps in parallel, improving execution times in all of them. In addition, our parallel approach includes a parallel Kmeans++ algorithm implementation and defines a new variant of Kmeans++ to reduce the impact of choosing outliers as initial centroids. The results show that our approach provides clustering quality results very similar to FFClust, and it requires an execution time of 3.5 s for processing about a million fibers, achieving a speedup of 11.5 times compared to FFClust.

A Novel Text Ensemble Clustering Based on Weighted Entropy Filtering Model

Text clustering is one of the important technical bases of natural language processing, and ensemble clustering improves the robustness of text clustering. According to the existing research of scholars and experts, the quality and diversity of basic clustering have a great influence on consensus clustering, and it has a particularly significant effect on text clustering. However, there are a few pieces of research aiming at reducing the number of low-quality clustering in ensembles. This paper proposes a novel clustering filtering model based on entropy criteria. The entropy criterion is used to evaluate the uncertainty of each cluster w.r.t. the ensemble. Two indexes are proposed on the basis of the uncertainty of cluster, namely, Clustering Trend Index (CTI) which indicates the contribution of each cluster w.r.t. basic clustering, and Cluster Consistency Index(CCI) which indicates the degree of cluster dispersion in the basic clustering. The proposed clustering filtering model is built on the basis of new weight using two proposed indexes. Thereby, by dropping the low-quality clustering, the percentage of high-quality clustering will increase. A large number of experiments on various real text data sets using optimal thresholds show that the proposed method has greatly improved accuracy and robustness, and is superior to existing ensemble clustering algorithms.

Urban Hotspot Area Detection Using Nearest-Neighborhood-Related Quality Clustering on Taxi Trajectory Data

Urban hotspot area detection is an important issue that needs to be explored for urban planning and traffic management. It is of great significance to mine hotspots from taxi trajectory data, which reflect residents’ travel characteristics and the operational status of urban traffic. The existing clustering methods mainly concentrate on the number of objects contained in an area within a specified size, neglecting the impact of the local density and the tightness between objects. Hence, a novel algorithm is proposed for detecting urban hotspots from taxi trajectory data based on nearest neighborhood-related quality clustering techniques. The proposed spatial clustering algorithm not only considers the maximum clustering in a limited range but also considers the relationship between each cluster center and its nearest neighborhood, effectively addressing the clustering issue of unevenly distributed datasets. As a result, the proposed algorithm obtains high-quality clustering results. The visual representation and simulated experimental results on a real-life cab trajectory dataset show that the proposed algorithm is suitable for inferring urban hotspot areas, and that it obtains better accuracy than traditional density-based methods.

Consensus clustering applied to multi-omics disease subtyping

BackgroundFacing the diversity of omics data and the difficulty of selecting one result over all those produced by several methods, consensus strategies have the potential to reconcile multiple inputs and to produce robust results.ResultsHere, we introduce ClustOmics, a generic consensus clustering tool that we use in the context of cancer subtyping. ClustOmics relies on a non-relational graph database, which allows for the simultaneous integration of both multiple omics data and results from various clustering methods. This new tool conciliates input clusterings, regardless of their origin, their number, their size or their shape. ClustOmics implements an intuitive and flexible strategy, based upon the idea of evidence accumulation clustering. ClustOmics computes co-occurrences of pairs of samples in input clusters and uses this score as a similarity measure to reorganize data into consensus clusters.ConclusionWe applied ClustOmics to multi-omics disease subtyping on real TCGA cancer data from ten different cancer types. We showed that ClustOmics is robust to heterogeneous qualities of input partitions, smoothing and reconciling preliminary predictions into high-quality consensus clusters, both from a computational and a biological point of view. The comparison to a state-of-the-art consensus-based integration tool, COCA, further corroborated this statement. However, the main interest of ClustOmics is not to compete with other tools, but rather to make profit from their various predictions when no gold-standard metric is available to assess their significance.AvailabilityThe ClustOmics source code, released under MIT license, and the results obtained on TCGA cancer data are available on GitHub: https://github.com/galadrielbriere/ClustOmics.

Scalable community detection via parallel correlation clustering

Graph clustering and community detection are central problems in modern data mining. The increasing need for analyzing billion-scale data calls for faster and more scalable algorithms for these problems. There are certain trade-offs between the quality and speed of such clustering algorithms. In this paper, we design scalable algorithms that achieve high quality when evaluated based on ground truth. We develop a generalized sequential and shared-memory parallel framework based on the LAMBDACC objective (introduced by Veldt et al.), which encompasses modularity and correlation clustering. Our framework consists of highly-optimized implementations that scale to large data sets of billions of edges and that obtain high-quality clusters compared to ground-truth data, on both unweighted and weighted graphs. Our empirical evaluation shows that this framework improves the state-of-the-art trade-offs between speed and quality of scalable community detection. For example, on a 30-core machine with two-way hyper-threading, our implementations achieve orders of magnitude speedups over other correlation clustering baselines, and up to 28.44× speedups over our own sequential baselines while maintaining or improving quality.

Spatiotemporal Data-Adaptive Clustering Algorithm: An Intelligent Computational Technique for City Big Data

The emerging paradigm of the modern city demands big data analytics in urban studies. Data mining techniques for big data have the potential to ascertain strategic insights for current and future cities. Clustering is increasingly growing in data mining, serving as a knowledge discovery tool. This article proposes a spatiotemporal clustering method, called spatiotemporal data-adaptive clustering (STDAC), to cope with clustering challenges, such as feature type, parameter setting, and treating spatial and temporal dimensions as equals. The proposed algorithm takes a dual-structure approach to obtain high-quality clusters and discover temporal changes not detected by other techniques. STDAC refers to the feature type and uses k-fold cross-validation (KCV) to replace the user-defined parameters with the data-driven values. The data-driven threshold from KCV performs as the endogenous variable, thus not requiring a priori assumptions or parameter settings. This article used the bus, taxi passengers, and de facto population data as illustrative case studies. The clustering performances were evaluated based on validation indexes, such as the Davies–Bouldin index (DBI) and Dunn index (DI), where STDAC generally had lower DBI and greater DI. The findings showed that STDAC could yield better clustering performances than other established algorithms. Hence, the proposed method would be promising to future studies requiring spatiotemporal big data analytics in urban studies, administrative management, and other fields.

Optimizing modularity with nonnegative matrix factorization

Community structure detection is one of the fundamental problems in complex network analysis towards understanding the topology structure and function of the network. Modularity is a criterion to evaluate the quality of community structures, and optimization of this quality function over the possible divisions of a network is a sensitive detection method for community structure. However, the direct application of this method is computationally costly. Nonnegative matrix factorization (NMF) is a widely used method for community detection. In this paper, we show that modularity maximization can be approximately reformulated under the framework of NMF with Frobenius norm, especially when [Formula: see text] is large. A new algorithm for detecting community structure is proposed based on the above finding. The new method is compared with four state-of-the-art methods on both synthetic and real-world networks, showing its higher clustering quality over the existing methods.

Leveraging the partition selection bias to achieve a high-quality clustering of mass spectra

In proteomics, the identification of peptides from mass spectral data can be mathematically described as the partitioning of mass spectra into clusters (i.e., groups of spectra derived from the same peptide). The way partitions are validated is just as important, having evolved side by side with the clustering algorithms themselves and given rise to many partition assessment measures. An assessment measure is said to have a selection bias if, and only if, the probability that a randomly chosen partition scoring a high value depends on the number of clusters in the partition. In the context of clustering mass spectra, this might mislead the validation process to favor clustering algorithms that generate too many (or few) spectral clusters, regardless of the underlying peptide sequence. A selection bias toward the number of peptides is desirable for proteomics as it estimates the number of peptides in a complex protein mixture. Here, we introduce an assessment measure that is purposely biased toward the number of peptide ion species. We also introduce a partition assessment framework for proteomics, called the Partition Assessment Tool, and demonstrate its importance by evaluating the performance of eight clustering algorithms on seven proteomics datasets while discussing the trade-offs involved. SignificanceClustering algorithms are widely adopted in proteomics for undertaking several tasks such as speeding up search engines, generating consensus mass spectra, and to aid in the classification of proteomic profiles. Choosing which algorithm is most fit for the task at hand is not simple as each algorithm has advantages and disadvantages; furthermore, specifying clustering parameters is also a necessary and fundamental step. For example, deciding on whether to generate “pure clusters” or fewer clusters but accepting noise. With this as motivation, we verify the performance of several widely adopted algorithms on proteomic datasets and introduce a theoretical framework for drawing conclusions on which approach is suitable for the task at hand.