Data Clustering Research Articles

GeoTemporal clustering for aquifer delineation: a big data approach to synchronizing and analyzing variable-length groundwater time series

Groundwater is a vital global resource. However, mapping aquifers remains challenging, particularly in developing nations. This study proposes a novel methodology for aquifer delineation using time-series clustering of groundwater-level data. The modular clustering framework utilizes hierarchical agglomerative clustering and a custom hydrology-specific distance function. This accounts for the variability in the length, temporal position, and consistency of the time series, in addition to gaps in records, aligning them temporally before comparison. Advantages over traditional techniques such as dynamic time warping, and Euclidean distance are provided for analyzing real-world hydrological data. The algorithm was optimized on a synthetic Texas aquifer dataset to identify the minimum time series lengths required for accurate clustering (> 90% accuracy). Applying this to real data from the Texas Groundwater Database GWDB with over one million readings and 60,000 wells, the modeling achieved ~ 73% accuracy, delineating the nine major Texan aquifers using a filtered number of 74 representative wells. The aquifer boundaries were geographically visualized using the GeoZ library. These findings suggest the effectiveness of groundwater characterization given the limited data. The optimized algorithm could provide inexpensive mapping capabilities in developing nations, requiring only historical data from existing wells over the decades. This technique is adaptive and can be improved through ongoing monitoring. The algorithm components are modular and upgradable thus future studies should optimize and test their generalizability using additional datasets.

Kernel-bounded clustering for spatial transcriptomics enables scalable discovery of complex spatial domains.

Spatial transcriptomics are a collection of technologies that have enabled characterization of gene expression profiles and spatial information in tissue samples. Existing methods for clustering spatial transcriptomics data have primarily focused on data transformation techniques to represent the data suitably for subsequent clustering analysis, often using an existing clustering algorithm. These methods have limitations in handling complex data characteristics with varying densities, sizes, and shapes (in the transformed space on which clustering is performed), and they have high computational complexity, resulting in unsatisfactory clustering outcomes and slow execution time even with GPUs. Rather than focusing on data transformation techniques, we propose a new clustering algorithm called kernel-bounded clustering (KBC). It has two unique features: (1) It is the first clustering algorithm that employs a distributional kernel to recruit members of a cluster, enabling clusters of varying densities, sizes, and shapes to be discovered, and (2) it is a linear-time clustering algorithm that significantly enhances the speed of clustering analysis, enabling researchers to effectively handle large-scale spatial transcriptomics data sets. We show that (1) KBC works well with a simple data transformation technique called the Weisfeiler-Lehman scheme, and (2) a combination of KBC and the Weisfeiler-Lehman scheme produces good clustering outcomes, and it is faster and easier-to-use than many methods that employ existing clustering algorithms and data transformation techniques.

ScSCC: A swapped contrastive learning‐based clustering method for single‐cell gene expression data

AbstractCell clustering plays a pivotal role in deciphering the intricacies of cell types, facilitating subsequent cell annotation endeavors within scRNA‐seq data analysis. In this paper, we propose a novel swapped contrastive clustering algorithm for scRNA‐seq data called scSCC. scSCC combines two contrastive learning modules, namely the instance contrastive learning module and the swapped prediction module, to extract clustering‐friendly cell representations. Through the combination of swapped prediction module and instance contrastive learning module, scSCC can retrieve disentangled cell representations and amplify the clustering signals in the latent space, leading to satisfactory clustering performance. Different from existing contrastive‐learning‐based scRNA‐seq data clustering algorithms, the swapped prediction module of scSCC injects clustering signals to the latent space through some clustering prototypes. The swapped prediction module encourages cells of the same cluster to gravitate toward the common clustering prototype and naturally stay away from other prototypes in the latent space, hence cell representations obtained by scSCC are more clustering‐friendly compared to other algorithms. Experimental results on real scRNA‐seq datasets show that scSCC achieves improved clustering performance compared with the benchmark methods. The ablation study on two contrastive modules exhibits the promotion by the combination of instance learning module and swapped prediction module. The source codes are available at the GitHub website (EnchantedJoy/scSCC).

Short-Term Building Electrical Load Prediction by Peak Data Clustering and Transfer Learning Strategy

With the gradual penetration of new energy generation and storage to the building side, the short-term prediction of building power demand plays an increasingly important role in peak demand response and energy supply/demand balance. The low occurring frequency of peak electrical loads in buildings leads to insufficient data sampling for model training, which is currently an important factor affecting the performance of short-term electrical load prediction. To address this issue, by using peak data clustering and knowledge transfer from similar buildings, a short-term electrical load forecasting method is proposed. First, a building’s electrical peak loads are clustered through peak/valley data analysis and K-nearest neighbors categorization method, thereby addressing the challenge of data clustering in data-sparse scenarios. Second, for peak/valley data clusters, an instance-based transfer learning (IBTL) strategy is used to transfer similar data from multi-source domains to enhance the target prediction’s accuracy. During the process, a two-stage similar data selection strategy is applied based on Wasserstein distance and locality sensitive hashing. An IBTL strategy, iTrAdaboost-Elman, is designed to construct the predictive model. The performance of proposed method is validated on a public dataset. Results show that the data clustering and transfer learning method reduces the error by 49.22% (MAE) compared to the Elman model. Compared to the same transfer learning model without data clustering, the proposed approach also achieves higher prediction accuracy (1.96% vs. 2.63%, MAPE). The proposed method is also applied to forecast hourly/daily power demands of two real campus buildings in the USA and China, respectively. The effects of data clustering and knowledge transfer are both analyzed and compared in detail.

SLAMF4 orchestrates the cytotoxic response of CD4+ T cells in rheumatoid arthritis.

This study aimed to assess whether signaling lymphocytic activation molecule family receptors (SLAMFs) are involved in the shaping of the pathological response of CD4+ T cells in rheumatoid arthritis (RA). Peripheral blood (PB) and synovial fluid (SF) mononuclear cells from RA patients were freshly isolated. In RA, we used a multi-modal approach to determine the involvement of numerous subpopulations of CD4+ T cells expressing SLAMFs. Experimentally, multiple flow cytometry panels, RNA sequencing and ex vivo stimulations were used. Analyses involved high-dimensional unsupervised clustering of flow cytometry data and pathway enrichment analyses of transcriptomic data. In PB of RA patients with active disease, SLAMF4+ effector memory T cells (Tem) represented the only overrepresented subpopulation of CD4+ T cells expressing SLAMFs. This positive correlation between RA activity and SLAMF4+ Tem was restricted to those co-expressing the intracellular molecule SAP and the tissue-homing receptor CCR5. Gene Set Enrichment Analysis of RNA-sequencing data reveals that SLAMF4+ CCR5+ Tem display a cytotoxicity-related gene signature. Moreover, based on the differential expression of cytotoxicity markers (GPR56, CX3CR1, granzyme-B, perforin and granulysin), unsupervised clustering of flow cytometry data identified distinct subpopulations of PB cytotoxic Tem. Among them, only SLAMF4high SAP+ CCR5+ Tem (Cytotox-F4high Tem) were correlated with RA activity. Remarkably, Cytotox-F4high Tem emerged as the only cytotoxic population of CD4+ T cells (CD4+ CTLs) present in SF of patients with active disease. This study emphasizes that Cytotox-F4high Tem represent a significant CD4+ CTL subpopulation involved in RA, suggesting that their inhibition represent a promising therapeutic interest.

Machine Learning-Driven Soft Sensor Implementation for Real-Time Fault Detection in CDU of Oil Refinery

Soft sensors in oil refineries provide operators with important insights into the behavior and performance of processes using real-time and historical data to generate predictions. This data-driven strategy makes it easier to make wise decisions for detecting faults, thus improving process optimization and control. The Crude Distillation Unit (CDU) imposes very harsh working environments for measuring instruments, imposing both the use of a very robust sensory system and periodic maintenance procedures, which are time-consuming and costly. Notwithstanding such precautions, faults in those measuring devices, such as temperature and pressure sensors, still occur, and the presence of a sensor fault deteriorates the efficiency, productivity, and reliability of the refinery process. Recent works focused only on some fault types (e.g., bias and drift), ignoring others. This study presents the design of a soft sensor to detect all possible fault types in the real-time processing of an oil refinery. This method used actual data collected from the Salahuddin oil refinery in Iraq, several preprocessing methods, and a machine-learning approach. The proposed soft sensor was designed using several stages, including data collection, preprocessing, clustering, and classification. In the classification stage, an approach based on a Bagged Decision Tree (BDT) and Support Vector Machine (SVM) was implemented to classify the detected faults. The proposed soft sensor was trained and tested using actual data, achieving a high fault detection and classification result of 99.96%.

An Energy-Efficient Hybrid LEACH Protocol that Enhances the Lifetime of Wireless Sensor Networks

A Wireless Sensor Networks (WSN) comprises of little, low-power sensors, which have a low battery limit and are often utilized in unfavorable conditions. These sensors are data processing and networking devices. The battery life, processing power, communication range, and memory of sensors, which are tiny, self-contained devices, are just a few of their limits. They also have transceivers, which gather environmental data and transmit it to a base station. There are various strategies accessible to expand the existence of a WSN and diminish energy utilization. LEACH (Low Energy Adaptive Clustering Hierarchy) is a successful and widely used method. The protocol clusters the network in order to conserve energy The cluster head, which is a single-region indicative node, receives fusion cluster data once the sensor nodes are divided into clusters. In our study, we provide a novel method for reducing sensor node energy consumption and extending network lifetime. Our research aims to increase network lifespan by reducing energy consumption based on the restrictions of the LEACH and its related algorithms.

IMPLEMENTASI K-MEANS CLUSTERING DALAM PENGELOMPOKAN DATA KUNJUNGAN WISATAWAN ASING DI INDONESIA

Clustering is a data mining technique used for grouping data based on specific similarities. This study implements K-Means Clustering to analyze foreign tourist visit data in Indonesia in 2024. Using the Knowledge Discovery in Database (KDD) methodology, the research involves five stages: Data Selection, preprocessing, Transformation, data mining, and Evaluation. Data Clustering was conducted using RapidMiner software, experimenting with different cluster counts (k=2 to k=7) to determine the optimal number of clusters. Results indicate that three clusters (k=3) with the smallest Davies-Bouldin Index (DBI) value were optimal. This Clustering approach categorizes tourists into low, medium, and high visit groups, assisting policymakers in strategic tourism development. The findings support capacity planning and seasonal marketing strategies to optimize Indonesia's tourism sector.

Genetic Diversity and Population Structure of Phyllosphere-Associated Xanthomonas euvesicatoria Bacteria in Physalis pubescens Based on BOX-PCR and ERIC-PCR in China

<i>Xanthomonas euvesicatoria</i> has become a serious problem in <i>Physalis pubescens</i>, leading to substantial crop losses. In our previous investigation, we used rapid molecular detection techniques to identify <i>X. euvesicatoria</i>; however, this pathogen’s diversity and population structure remain poorly understood, despite their importance in disease management. To address this knowledge gap, we analyzed the diversity of <i>X. euvesicatoria</i> using BOX-PCR and ERIC-PCR fingerprinting techniques. A total of 103 isolates were collected from 13 counties across Heilongjiang province during the 2018 and 2019 growing seasons. Our findings revealed 635 unique genetic patterns from ERIC-PCR fingerprinting, compared to 360 patterns from BOX-PCR. BOX-PCR analysis identified 12 distinct genotypic clusters, whereas ERIC-PCR identified 14 clusters through unweighted pair group approach with arithmetic average analysis, demonstrating substantial genetic variability. STRUCTURE analysis further identified five distinct genetic clusters in the BOX-PCR data and two in the ERIC-PCR data. The Hailin isolates showed the highest level of diversification compared to other regional isolates. AMOVA results indicated that 85% of the genetic variation in BOX-PCR was attributable to within-population differences, while 78% of ERIC-PCR variation was due to differences across populations. In addition, a Mantel test demonstrated a tenuous correlation between BOX-PCR and ERIC-PCR genetic markers, indicating distinct genetic profiles. This extensive genetic information enhances our understanding of the epidemiology of bacterial leaf spot and its potential therapeutic prospects. These data can provide insights into <i>Xanthomonas</i> strains’ diversity and geographical dissemination.

Non-centroid-based discrete differential evolution for data clustering

Non-centroid-based discrete differential evolution for data clustering

Determination of biomass energy potential based on regional characteristics using adaptive clustering method

Determining the energy potential of biomass is the first step in selecting the most suitable and efficient energy conversion technology based on regional characteristics. The approach to estimating and determining biomass potential generally uses geospatial technology related to collecting and processing data about mapping an area. Unfortunately, this method is inadequate for simulating the interaction between variables, nor can it provide accurate predictions for the biomass supply chain. As a result, the results obtained from this method tend to be biased and macro, particularly in regions experiencing rapid land-use development. In this paper, the author has developed a clustering methodology with a fuzzy c-means (FCM) algorithm to determine biomass energy potential based on regional characteristics to produce data clusters with high accuracy. Grouping the characteristics of clustering-based areas involves grouping physical or abstract objects into classes or similar objects.

Automated Machine Learning of Interfacial Interaction Descriptors and Energies in Metal-Catalyzed N2 and CO2 Reduction Reactions.

The applications of machine learning (ML) in complex interfacial interactions are hindered by the time-consuming process of manual feature selection and model construction. An automated ML program was implemented with four subsequent steps: data distribution analysis, dimensionality reduction and clustering, feature selection, and model optimization. Without the need of manual intervention, the descriptors of metal charge variance (ΔQCT) and electronegativity of substrate (χsub) and metal (δχM) were raised up with good performance in predicting electrochemical reaction energies for both nitrogen reduction reaction (NRR) and CO2 reduction reaction (CO2RR) on metal-zeolites and MoS2 surfaces. The important role of interfacial interactions in tuning the catalytic reactivity in NRR and CO2RR was highlighted from SHAP analysis. It was proposed that Fe-, Cr-, Zn-, Nb-, and Ta-zeolites are favorable catalysts for NRR, while Ni-zeolite showed a preference for CO2RR. An elongated bond of N2 or a bent configuration of CO2 was shown in V-, Co-, and Mo-zeolites, indicating that the molecule could be activated after the adsorption in both NRR and CO2RR pathways. The generalizability of the automatically built ML model is demonstrated from applications to other catalytic systems such as metal-organic frameworks and SiO2 surfaces. The automated ML program is a useful tool to accelerate the data-driven exploration of relationship between structures and material properties without the need of manual feature selection.

Optimizing K-Means Algorithm Using the Purity Method for Clustering Oil Palm Producing Regions in North Aceh

The K-Means algorithm is a fundamental tool in machine learning, widely utilized for data clustering tasks. This research aims to improve the performance of the K-Means algorithm by integrating the Purity method, specifically focusing on clustering regions renowned for oil palm production in North Aceh. Oil palm cultivation is a vital agricultural sector in North Aceh, contributing significantly to the local economy and employment. This study examines two clustering techniques: the conventional K-Means algorithm and an optimized version, Purity+K-Means. The integration of the Purity method increases the efficiency of K-Means by decreasing the required iterations for convergence. The data used for clustering analysis is sourced from the Department of Agriculture and Food in North Aceh Regency and pertains to oil palm production in 2023. The findings indicate that the Purity+K-Means approach notably reduces the iteration count and improves cluster quality. The average Davies-Bouldin Index (DBI) for standard K-Means is 0.45, whereas the Purity+K-Means method lowers it to 0.30. Furthermore, applying the Purity method reduced the number of K-Means iterations from 15 to just 3. These results highlight an enhancement in clustering performance and overall efficiency.

Eye tracking based clustering using the Korean version of the reading the mind in the eyes test

This study applies eye-tracking paradigms to cluster data based on participants’ gaze patterns, while performing the Korean version of the Reading the Mind in Eyes Test, and to investigate whether there were differences in the neurocognitive and other Theory of Mind (ToM) tests among the classified clusters. A total of 89 (50 males) non-clinical youths were recruited. The k-means algorithm was adopted, and the optimised number of clusters was determined using the elbow, silhouette and NbClust methods. Furthermore, multivariate analysis was employed to determine whether there were differences among the clusters in the neurocognitive and other ToM tests. Four clusters were proposed based on the index used to estimate the optimised cluster. The long word fixation time cluster had significantly more total errors and fewer categories completed in the Wisconsin Card Sorting Test, lower backward Digit Span score, and lower sequencing scores in the Theory of Mind Picture Stories Task than in the other clusters. The main findings suggested that even when performing a perceptual-level ToM task that requires the ability to understand mental states, at least in some individuals, gaze patterns are related to neurocognitive strategies, especially executive function, rather than to the specific social cognitive function itself.

Integration of field data and application of machine learning methods to assess the condition of the near-wellbore zone of carbonate reservoirs

Relevance. Increase in the share of hard-to-recover reserves. Often, for the effective development of complex reservoirs, methods of enhanced oil recovery and production intensification are used. Currently, the feasibility of carrying out geological and technical measures is based on the results of interpretation well tests, which allows assessing the condition of the near-wellbore zone. The disadvantages of this research method are a long shutdown (as a result, “shortfalls” of oil) and increased risks of failure to bring wells into operation. In this regard, the integration of field data and the use of machine learning to describe the state of the near-wellbore zone can have a positive effect on the timeliness of geological and technical activities and ensure maximization of their effectiveness in the future. Aim. To develop a methodology for increasing the accuracy of the near-wellbore zone permeability prediction of carbonate reservoirs based on the use of machine learning methods. Methods. Statistical methods, solving the classification problem using machine learning methods. Results. This paper proposes an approach for quickly assessing the permeability of the near-wellbore zone, based on a statistical analysis of the results of interpretation of hydrodynamic studies (256 studies) and operational data from wells of an oil carbonate reservoir in the Perm Krai. To assess near-wellbore zone permeability, a multiple linear regression model was built. In order to improve the statistical metrics of regression of the near-wellbore zone permeability, the dependence of this parameter on the specific productivity coefficient in the conditions of a carbonate reservoir was studied and divided into clusters. The SHAP library was used to identify significant parameters on the predicted value. To perform the task of classifying clusters based on source data, the authors have used a machine learning technique – support vector machine and constructed differentially the regression models for each cluster. Using of this approach made it possible to increase the coefficient of determination from 0.76 to 0.96 and reduce the average absolute error in predicting the near-wellbore zone permeability from 0.018 to 0.007 µm2. Thus, the authors proposed a methodology for predicting the near-wellbore zone permeability using statistical methods based on preliminary clustering of the initial data and their classification using machine learning approaches.

Using Hybrid LSTM Neural Networks to Detect Anomalies in the Fiber Tube Manufacturing Process

The production process of tubes for fiber optic cables is a complex process, where proper execution is crucial to the quality of the final product. This process has a complex state vector whose structure and coordinates dynamically change during the tube extrusion process. Small fluctuations in process parameters, such as temperature, extrusion pressure, production speed, and optical fiber tension, affect the optical attenuation of the final product. Such defects necessitate the withdrawal of the product. Due to the high number of process coordinates and the technological inability to automatically label those segments of the production process that cause anomalies in the final product, the authors used data clustering methods to create a training set that enabled the use of neural tools for anomaly detection. The system proposed in the main part of the paper includes a hybrid Long short-term memory (LSTM) network model, which is fed with data streams recorded on the tube extrusion production line. The input module, which performs preprocessing of input data, conducts multiresolution analysis of recorded process parameters, and recommends the process state’s belonging to a set of classes describing individual production anomalies to appropriate LSTM network modules. The learning process of the three–channel network allowed effective recognition of five classes of the monitored tube production process. The fit level of the proposed network model reached R2 values of ≥0.85.

ScHNTL: single-cell RNA-seq data clustering augmented by high-order neighbors and triplet loss.

The rapid development of single-cell RNA sequencing (scRNA-seq) has significantly advanced biomedical research. Clustering analysis, crucial for scRNA-seq data, faces challenges including data sparsity, high dimensionality, and variable gene expressions. Better low-dimensional embeddings for these complex data should maintain intrinsic information while making similar data close and dissimilar data distant. However, existing methods utilizing neural networks typically focus on minimizing reconstruction loss and maintaining similarity in embeddings of directly related cells, but fail to consider dissimilarity, thus lacking separability and limiting the performance of clustering. We propose a novel clustering algorithm, called scHNTL (scRNA-seq data clustering augmented by high-order neighbors and triplet loss). It first constructs an auxiliary similarity graph and uses a Graph Attentional Autoencoder to learn initial embeddings of cells. Then it identifies similar and dissimilar cells by exploring high-order structures of the similarity graph and exploits a triplet loss of contrastive learning, to improve the embeddings in preserving structural information by separating dissimilar pairs. Finally, this improvement for embedding and the target of clustering are fused in a self-optimizing clustering framework to obtain the clusters. Experimental evaluations on 16 real-world datasets demonstrate the superiority of scHNTL in clustering over the state-of-the-arts single-cell clustering algorithms. Python implementation of scHNTL is available at Figshare (https://doi.org/10.6084/m9.figshare.27001090) and Github (https://github.com/SWJTU-ML/scHNTL-code). Supplementary data are available at Bioinformatics online.

ScSMD: a deep learning method for accurate clustering of single cells based on auto-encoder

BackgroundSingle-cell RNA sequencing (scRNA-seq) has transformed biological research by offering new insights into cellular heterogeneity, developmental processes, and disease mechanisms. As scRNA-seq technology advances, its role in modern biology has become increasingly vital. This study explores the application of deep learning to single-cell data clustering, with a particular focus on managing sparse, high-dimensional data.ResultsWe propose the SMD deep learning model, which integrates nonlinear dimensionality reduction techniques with a porous dilated attention gate component. Built upon a convolutional autoencoder and informed by the negative binomial distribution, the SMD model efficiently captures essential cell clustering features and dynamically adjusts feature weights. Comprehensive evaluation on both public datasets and proprietary osteosarcoma data highlights the SMD model’s efficacy in achieving precise classifications for single-cell data clustering, showcasing its potential for advanced transcriptomic analysis.ConclusionThis study underscores the potential of deep learning-specifically the SMD model-in advancing single-cell RNA sequencing data analysis. By integrating innovative computational techniques, the SMD model provides a powerful framework for unraveling cellular complexities, enhancing our understanding of biological processes, and elucidating disease mechanisms. The code is available from https://github.com/xiaoxuc/scSMD.

Learning model combined with data clustering and dimensionality reduction for short-term electricity load forecasting

Electric load forecasting is crucial in the planning and operating electric power companies. It has evolved from statistical methods to artificial intelligence-based techniques that use machine learning models. In this study, we investigate short-term load forecasting (STLF) for large-scale electricity usage datasets. We propose a new prediction model for STLF that combines data clustering and dimensionality reduction schemes to handle large-scale electricity usage data effectively. Here, we adapt k-means clustering for data clustering, kernel principal component analysis (kernel PCA), universal manifold approximation and projection (UMAP), and t-stochastic nearest neighbor (t-SNE) for dimensionality reduction. To verify the effectiveness of the proposed model, we extensively apply it to neural network-based models. We compare and analyze the performance of the proposed model with the comparisons using actual electricity usage data for 4710 households. Experimental results demonstrate that data clustering with dimensionality reduction can improve the performance of baseline models. As a result, the prediction accuracy of the proposed method outperforms those of the existing methods by 1.01–1.76 times for summer data and by 1.03–1.36 times for winter data in terms of mean absolute percentage error (MAPE).

Artificial Afterimage Algorithm: A New Bio-Inspired Metaheuristic Algorithm and Its Clustering Application

Metaheuristic methods are optimization methods that look for different ways to converge to a solution to a problem where it is difficult to find a solution analytically. Their difference from known optimization methods is that they imitate living things or systems in nature. Each metaheuristic method has its equations, and the solution is found using these equations. In this study, a new, metaheuristic method called the afterimage algorithm is proposed. The proposed method was developed inspired by the fact that when we close our eyes after looking at a luminous image for a while, the vision still occurs in our minds. This is called an afterimage. The proposed method first pre-processes with the operator called afterimage and calculates the best and worst solution values. The visual angle value is then calculated, and new solutions are produced around this value. Three different datasets were used in experimental studies on data clustering. Accuracies of 96.66% for the iris plant dataset, 92% for the Wisconsin breast cancer dataset, and 95% for the occupancy detection dataset were obtained.