Clustering Accuracy Research Articles

Phase-Angle-Encoded Snake Optimization Algorithm for K-Means Clustering

The rapid development of metaheuristic algorithms proves their advantages in optimization. Data clustering, as an optimization problem, faces challenges for high accuracy. The K-means algorithm is traditaaional but has low clustering accuracy. In this paper, the phase-angle-encoded snake optimization algorithm (θ-SO), based on mapping strategy, is proposed for data clustering. The disadvantages of traditional snake optimization include slow convergence speed and poor optimization accuracy. The improved θ-SO uses phase angles for boundary setting and enables efficient adjustments in the phase angle vector to accelerate convergence, while employing a Gaussian distribution strategy to enhance optimization accuracy. The optimization performance of θ-SO is evaluated by CEC2013 datasets and compared with other metaheuristic algorithms. Additionally, its clustering optimization capabilities are tested on Iris, Wine, Seeds, and CMC datasets, using the classification error rate and sum of intra-cluster distances. Experimental results show θ-SO surpasses other algorithms on over 2/3 of CEC2013 test functions, hitting a 90% high-performance mark across all clustering optimization tasks. The method proposed in this paper effectively addresses the issues of data clustering difficulty and low clustering accuracy.

Robust Parameter Optimisation of Noise-Tolerant Clustering for DENCLUE Using Differential Evolution

Clustering samples based on similarity remains a significant challenge, especially when the goal is to accurately capture the underlying data clusters of complex arbitrary shapes. Existing density-based clustering techniques are known to be best suited for capturing arbitrarily shaped clusters. However, a key limitation of these methods is the difficulty in automatically finding the optimal set of parameters adapted to dataset characteristics, which becomes even more challenging when the data contain inherent noise. In our recent work, we proposed a Differential Evolution-based DENsity CLUstEring (DE-DENCLUE) to optimise DENCLUE parameters. This study evaluates DE-DENCLUE for its robustness in finding accurate clusters in the presence of noise in the data. DE-DENCLUE performance is compared against three other density-based clustering algorithms—DPC based on weighted local density sequence and nearest neighbour assignment (DPCSA), Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and Variable Kernel Density Estimation–based DENCLUE (VDENCLUE)—across several datasets (i.e., synthetic and real). The study has consistently shown superior results for DE-DENCLUE compared to other models for most datasets with different noise levels. Clustering quality metrics such as the Silhouette Index (SI), Davies–Bouldin Index (DBI), Adjusted Rand Index (ARI), and Adjusted Mutual Information (AMI) consistently show superior SI, ARI, and AMI values across most datasets at different noise levels. However, in some cases regarding DBI, the DPCSA performed better. In conclusion, the proposed method offers a reliable and noise-resilient clustering solution for complex datasets.

Research on a Multi-Scale Clustering Method for Buildings Taking into Account Visual Cognition

Building clustering is a key problem that needs to be solved in the realization of the automatic synthesis of large-scale maps. The selection of different feature and spatial distance calculation methods has a great impact on the clustering results, and the need to manually select appropriate feature and distance metrics leads to the problem of not being able to fully consider the complexity and diversity of buildings. In this paper, we propose a multi-scale clustering method for buildings that takes visual perception into account using the Gestalt principle to simulate how humans classify buildings through visual perception. Moreover, by analyzing the spatial features and texture attributes of buildings, a visual distance is designed to be used as a condition for building classification to assess the similarity between buildings, solving the complexity of manually selecting feature vectors and spatial distances and realizing the adaptive selection of features. Through experimental validation at different scales (macro, meso and micro), the present method is able to achieve the accurate clustering of buildings, and a frequency threshold of 91% is found, which is able to determine the optimal clustering results. The experimental results show that the proposed method can not only fully consider the complexity and diversity of buildings but also effectively support the understanding and analysis of urban spatial structure and provide a scientific decision-making basis for urban planning and management.

A Method for Predicting the Impact of Labor Education at Institutions of Higher Education Based on a Two-Stage Clustering Algorithm

In order to develop more accurate teaching plans and methods based on students’ characteristics and needs, and improve the effectiveness of labor education, a two-stage clustering algorithm based method for predicting the effectiveness of labor education in universities is studied. A two-stage clustering algorithm-based method for predicting the effectiveness of labor education in universities has been proposed. Through in-depth analysis of labor education data in universities, principal component analysis (PCA) was used to select 16 most representative variables from the original features. Subsequently, self-organizing map (SOM) neural network was used for preliminary clustering, and the optimal number of clusters was determined to be 8. Furthermore, the K-means clustering algorithm performs fine clustering based on the initial centers provided by SOM, significantly improving the stability and accuracy of clustering. At the same time, processing the outliers in the clustering results has reduced their impact on the clustering effect. Finally, the processed data was input into an improved XGBoost prediction model, which achieved a true positive rate (TPR) of over 85% on the ROC curve when predicting the effectiveness of labor education in universities. While maintaining a low false positive rate (FPR), the model outperformed other comparison methods significantly. This achievement not only validates the effectiveness of the proposed method in feature selection, clustering optimization, and prediction model construction, but also confirms its ability to accurately predict the labor education effectiveness of students in different grades, providing strong support for higher education institutions to optimize teaching plans and improve educational outcomes.

MitoSort: Robust Demultiplexing of Pooled Single-cell Genomics Data Using Endogenous Mitochondrial Variants.

Multiplexing across donors has emerged as a popular strategy to increase throughput, reduce costs, overcome technical batch effects, and improve doublet detection in single-cell genomic studies. To eliminate additional experimental steps, endogenous nuclear genome variants are used for demultiplexing pooled single-cell RNA sequencing (scRNA-seq) data by several computational tools. However, these tools have limitations when applied to single-cell sequencing methods that do not cover nuclear genomic regions well, such as single-cell assay for transposase-accessible chromatin with sequencing (scATAC-seq). Here, we demonstrate that mitochondrial germline variants are an alternative, robust, and computationally efficient endogenous barcode for sample demultiplexing. We propose MitoSort, a tool that uses mitochondrial germline variants to assign cells to their donor of origin and identify cross-genotype doublets in single-cell genomics datasets. We evaluate its performance by using in silico pooled mitochondrial scATAC-seq (mtscATAC-seq) libraries and experimentally multiplexed data with cell hashtags. MitoSort achieves high accuracy and efficiency in genotype clustering and doublet detection for mtscATAC-seq data, addressing the limitations of current computational techniques tailored for scRNA-seq data. Moreover, MitoSort exhibits versatility and can be applied to various single-cell sequencing approaches beyond mtscATAC-seq, provided the mitochondrial variants are reliably detected. Furthermore, we demonstrate the application of MitoSort in a case study where B cells from eight donors were pooled and assayed by single-cell multi-omics sequencing. Altogether, our results demonstrate the accuracy and efficiency of MitoSort, which enables reliable sample demultiplexing in various single-cell genomic applications. MitoSort is available at https://github.com/tangzhj/MitoSort.

Joint Beamforming Design and User Clustering Algorithm in NOMA-Assisted ISAC Systems.

To enhance the performance of non-orthogonal multiple access (NOMA)-assisted integrated sensing and communication (ISAC) systems in multi-user distributed scenarios, an improved Gaussian Mixture Model (GMM)-based user clustering algorithm is proposed. This algorithm is tailored for ISAC systems, significantly improving bandwidth reuse gains and reducing serial interference. First, using the Sum of Squared Errors (SSE), the algorithm reduces sensitivity to the initial cluster center locations, improving clustering accuracy. Then, direction weight factors are introduced based on the base station position and a penalty function involving users' Euclidean distances and sensing power. Modifications to the EM algorithm in calculating posterior probabilities and updating the covariance matrix help align user clusters with the characteristics of NOMAISAC systems. This improves users' interference resistance, lowers decoding difficulty, and optimizes the system's sensing capabilities. Finally, a fractional programming (FP) approach addresses the non-convex joint beamforming design problem, enhancing power and channel gains and achieving co-optimizing sensing and communication signals. The simulation results show that, under the improved GMM user clustering algorithm and FP optimization, the NOMA-ISAC system improves user spectral efficiency by 4.3% and base station beam intensity by 5.4% compared to traditional ISAC systems.

Uncertain data density peak clustering algorithm based on JS divergence

Aiming at the defects of traditional density-based uncertainty clustering algorithms, such as parameter sensitivity and poor clustering results for complex manifold uncertain data sets, a new uncertainty data density peak clustering algorithm based on JS divergence (UDPC-JS) is proposed. The algorithm first removes noise points by using the uncertain natural neighbor density factor defined by the uncertain natural neighbor. Secondly, the local density of uncertain data objects is calculated by combining the uncertain natural neighbor and JS divergence. The initial clustering center of the uncertain data set is found by combining the idea of representative points, and the distance between the initial clustering centers is defined based on JS divergence and graph. Then, the decision graph is constructed on the initial clustering center using the local density calculated based on the uncertain natural neighbor and JS divergence and the newly defined distance between the initial clustering centers based on JS divergence and graph, and the final clustering center is selected according to the decision graph. Finally, the unassigned uncertain data objects are assigned to the cluster where their initial clustering center is located. Experimental results show that the algorithm has better clustering effect and accuracy than the comparison algorithm, and has a greater advantage in processing complex manifold uncertain data sets.

DPC clustering algorithm based on K-nearest neighbors and kernel density estimation

Clustering by fast search and find of density peaks (DPC) is a density-based clustering algorithm. This algorithm does not require iteration and too many parameter settings, but it cannot identify cluster centers with small cluster density because the local structure of the data is not considered when calculating the local density. To address the above problems, a DPC clustering algorithm (KKDPC) based on K-reciprocal Neighbors (KN) and Kernel Density Estimation (KDE) is proposed. First, the number of reciprocal neighbors and local kernel density of data points are obtained by nearest neighbor and kernel density estimation methods ; second, the number of K reciprocal neighbors is added to the local kernel density to obtain a new local density; finally, the relative distance is obtained according to the local density of the data points, and the cluster center is selected and the non-center point is allocated by constructing a decision graph. Experiments are conducted using artificial and real data sets, and compared with seven clustering algorithms: DPC, DBSCAN, K-means, FCM, DPC-KNN, DPC-NN, and DPC-FWSN. The performance of the KKDPC clustering algorithm is verified by calculating the adjusted mutual information (AMI), adjusted Rand index (ARI), and normalized mutual information (NMI). The experimental results show that this method can more accurately identify the cluster center and effectively improve the clustering accuracy.

A clustering ensemble algorithm for handling deep embeddings using cluster confidence

Abstract Clustering ensemble, which aims to learn a robust consensus clustering from multiple weak base clusterings, has achieved promising performance on various applications. With the development of big data, the scale and complexity of data is constantly increasing. However, most existing clustering ensemble methods typically employ shallow clustering algorithms to generate base clusterings. When confronted with high-dimensional complex data, these shallow algorithms fail to fully utilize the intricate features present in the latent data space. As a result, the quality and diversity of the generated base clusterings are insufficient, thus affecting the subsequent ensemble performance. To address this issue, we propose a novel clustering ensemble algorithm for handling deep embeddings using cluster confidence (CEDECC) to improve the robustness and performance. Instead of simply combining deep clustering with clustering ensembles, we take into consideration that the performance of existing deep clustering methods heavily relies on the quality of low-dimensional embeddings generated during the pre-training stage. The quality of embeddings is unstable due to the influence of different initialization parameters. In CEDECC, specifically, we first construct a cluster confidence measure to evaluate the quality of low-dimensional embeddings. Typically, high-quality low-dimensional embeddings yield accurate clustering results with the same model parameters. Then, we utilize multiple high-quality embeddings to generate the base partitions. In the ensemble strategy phase, we consider the cluster-wise diversity and propose a novel ensemble cluster estimation to improve the overall consensus performance of the model. Extensive experiments on three benchmark datasets and four real-world biological datasets have demonstrated that the proposed CEDECC consistently outperforms the state-of-the-art clustering ensemble methods.

Research on high voltage circuit breaker multi-sensor signal feature selection method based on GA-Kmeans algorithm

Abstract Currently, in the research on fault information of high-voltage circuit breakers (HVCBs) based on multi-sensor data, issues such as shallow feature assessment and unclear feature selection rules exist, leading to a decrease in fault identification accuracy due to redundant characteristics. To address this, this paper proposes a novel feature selection method based on GA-Kmeans. This method encodes the original feature space into binary format and employs the clustering accuracy of the Kmeans model as the fitness function to obtain a high-quality feature subset that maximally represents and distinguishes faults. By combining different feature selection methods with fault diagnosis models, six typical application scenarios are constructed. Results indicate that compared to traditional methods such as Relief-F, KPCA, GA-SVM for feature selection, and SVM for fault diagnosis, the proposed GA-Kmeans method reduces the dimensionality of the original feature space and employs the Kmeans clustering algorithm as the diagnostic model, achieving a final diagnostic accuracy of 95.29%. This method outperforms others, with a 37.24% higher diagnostic accuracy than SVM under the original feature space, and a decrease of 47.90% in standard deviation. This validates the necessity of feature selection and the superiority of the proposed method, providing a reliable and stable diagnostic basis for subsequent mechanical fault diagnosis of HVCBs.

Enhanced tensor based embedding anchor learning for multi-view clustering

Existing anchor graph based multi-view clustering methods can overcome the problem of high computational cost in traditional multi-view clustering methods. However, the anchor points selected from high-dimensional data often contain irrelevant noise and outliers, which would affect the clustering performance. To address this issue, we propose an embedding anchor based multi-view clustering method, called enhanced tensor based embedding anchor learning (ETEAL). Specifically, we unify the learning process of latent embedding space, anchor points, and anchor graphs into a common framework, which eliminates noise and redundant data in the original space and enhances the synergistic optimization between the components. Meanwhile, we employ an enhanced tensor strategy to constrain the embedding anchor graphs, which exploits the higher-order relationships between views and recovers the global low-rank property of the embedding anchor graphs. Finally, we develop an anchor graph fusion strategy, which significantly reduces the huge overhead of traditional graph fusion that requires the construction of complete graphs. Experimental results on eight benchmark datasets show that the proposed method significantly outperforms other state-of-the-art methods in terms of scalability and clustering accuracy.

Spatially Informed Graph Structure Learning Extracts Insights from Spatial Transcriptomics.

Embeddings derived from cell graphs hold significant potential for exploring spatial transcriptomics (ST) datasets. Nevertheless, existing methodologies rely on a graph structure defined by spatial proximity, which inadequately represents the diversity inherent in cell-cell interactions (CCIs). This study introduces STAGUE, an innovative framework that concurrently learns a cell graph structure and a low-dimensional embedding from ST data. STAGUE employs graph structure learning to parameterize and refine a cell graph adjacency matrix, enabling the generation of learnable graph views for effective contrastive learning. The derived embeddings and cell graph improve spatial clustering accuracy and facilitate the discovery of novel CCIs. Experimental benchmarks across 86 real and simulated ST datasets show that STAGUE outperforms 15 comparison methods in clustering performance. Additionally, STAGUE delineates the heterogeneity in human breast cancer tissues, revealing the activation of epithelial-to-mesenchymal transition and PI3K/AKT signaling in specific sub-regions. Furthermore, STAGUE identifies CCIs with greater alignment to established biological knowledge than those ascertained by existing graph autoencoder-based methods. STAGUE also reveals the regulatory genes that participate in these CCIs, including those enriched in neuropeptide signaling and receptor tyrosine kinase signaling pathways, thereby providing insights into the underlying biologicalprocesses.

An Integer Programming Approach to Subspace Clustering with Missing Data

In the subspace clustering with missing data (SCMD) problem, we are given a collection of n partially observed d-dimensional vectors. The data points are assumed to be concentrated near a union of low-dimensional subspaces. The goal of SCMD is to cluster the vectors according to their subspace membership and recover the underlying basis, which can then be used to infer their missing entries. State-of-the-art algorithms for SCMD can fail on instances with a high proportion of missing data, with full-rank data, or if the underlying subspaces are similar to each other. We propose a novel integer programming approach for SCMD. The approach is based on dynamically determining a set of candidate subspaces and optimally assigning points to selected subspaces. The problem structure is identical to the classical facility-location problem, with subspaces playing the role of facilities and data points playing that of customers. We propose a column-generation approach for identifying candidate subspaces combined with a Benders decomposition approach for solving the linear programming relaxation of the formulation. An empirical study demonstrates that the proposed approach can achieve better clustering accuracy than state-of-the-art methods when the data are high rank, the percentage of missing data is high, or the subspaces are similar. Funding: Support for this research was provided by American Family Insurance through a research partnership with the University of Wisconsin–Madison’s Data Science Institute. Supplemental Material: The online appendix is available at https://doi.org/10.1287/ijoo.2023.0020 .

Active Fire Clustering and Spatiotemporal Dynamic Models for Forest Fire Management

Forest fires are increasingly destructive, contributing to significant ecological damage, carbon emissions, and economic losses. Monitoring these fires promptly and accurately, particularly by delineating fire perimeters, is critical for mitigating their impact. Satellite-based remote sensing, especially using active fire products from VIIRS and MODIS, has proven indispensable for real-time forest fire monitoring. Despite advancements, challenges remain in accurately clustering and delineating fire perimeters in a timely manner, as many existing methods rely on manual processing, resulting in delays. Active fire perimeter (AFP) and Timely Active Fire Progression (TAFP) models were developed which aim to be an automated approach for clustering active fire data points and delineating perimeters. The results demonstrated that the combined dataset achieved the highest matching rate of 85.13% for fire perimeters across all size classes, with a 95.95% clustering accuracy for fires ≥100 ha. However, the accuracy decreased for smaller fires. Overall, 1500 m radii with alpha values of 0.1 were found to be the most effective for fire perimeter delineation, particularly when applied at larger radii. The proposed models can play a critical role in improving operational responses by fire management agencies, helping to mitigate the destructive impact of forest fires more effectively.

Robust and stochastic sparse subspace clustering

Sparse subspace clustering (SSC) has been widely employed in machine learning and pattern recognition, but it still faces scalability challenges when dealing with large-scale datasets. Recently, stochastic SSC (SSSC) has emerged as an effective solution by leveraging the dropout technique. However, SSSC cannot robustly handle noise, especially non-Gaussian noise, leading to unsatisfactory clustering performance. To address the above issues, we propose a novel robust and stochastic method called stochastic sparse subspace clustering with the Huber function (S3CH). The key idea is to introduce the Huber surrogate to measure the loss of the stochastic self-expression framework, thus S3CH inherits the advantage of the stochastic framework for large-scale problems while mitigating sensitivity to non-Gaussian noise. In algorithms, an efficient proximal alternating minimization (PAM)-based optimization scheme is developed. In theory, the convergence of the generated sequence is rigorously proved. Extensive numerical experiments on synthetic and six real datasets validate the advantages of the proposed method in clustering accuracy, noise robustness, parameter sensitivity, post-hoc analysis, and model stability.

Clustering dynamic networks by discriminating roles of vertices and capturing temporality with subsequent feature projection

Clustering dynamic networks has gained popularity due to the need to analyze complex systems that evolve over time, which cannot be fully characterized by traditional static models. It is highly non-trivial in comparison to clustering static network since it requires simultaneously to balance clustering accuracy and clustering drift, where clustering accuracy measures how clustering reflects structure of graph at current time, and clustering drift quantifies how clustering smoothes historical snapshot(s). In this study, we propose an algorithm clustering dynamic network by discriminating roles of vertices and capturing temporality with subsequent feature projection (CDN-DRCT). Specifically, clustering accuracy is achieved by factorizing high-order matrix of slice at current time, and vertices are divided into static and dynamic ones by the reconstruction errors. Finally, the proposed algorithm measures temporality of networks with a projection matrix, which connects subsequent features at the previous and current time, thereby enhancing clustering drift of clusters. In this case, temporality of dynamic networks is characterized from vertex and global level, providing a better way to balance clustering accuracy and clustering drift. Experimental results on 10 typical dynamic networks demonstrate the proposed algorithm is superior to baselines in terms of accuracy as well efficiency.

Application of Improved HF-DBSCAN Algorithm in Analyzing Complex Trajectory Data of Wi-Fi Users in Smart Campus

ABSTRACT With the development of wireless network technology and the increasingly widespread application of mobile intelligent terminals, a large amount of wireless network data has been generated in smart campuses. By analyzing the trajectory data through clustering algorithms, more valuable user data information can be obtained. However, current clustering algorithms have high computational difficulty, long running time, and poor clustering accuracy for complex trajectories. Based on this background, the study proposes a spatial clustering of Application with Noise by Density-Based Spatial Clustering of Application with Noise combining Hausdorff and Frechet. The method is denoted as (Hausdorff and Frechet-Density Based Spatial Clustering of Application with Noise, HF-DBSCAN). DBSCAN denotes the Density Based Spatial Clustering of Application with Noise. In the experimental results, compared to the AH-DBSCAN, DBSCAN, and FD-DBSCAN algorithms, the running time of HF-DBSCAN algorithm in complex trajectory cluster analysis is reduced by 59.33%, 39.82%, and 35.12%, respectively, and the contour coefficient is closer to 1; the Davies-Boldin Index (DBI) values decreased by 59.24%, 64.42%, and 68.24%, respectively. The experiment shows that the optimized HF-DBSCAN algorithm has lower computational difficulty, better clustering performance, and higher clustering accuracy, which verifies the effectiveness of this study.

An Efficient Method for Identifying Inter-Well Connectivity Using AP Clustering and Graphical Lasso: Validation with Tracer Test Results

Identifying inter-well connectivity is crucial for optimizing reservoir development and facilitating informed adjustments. While current engineering methods are effective, they are often prohibitively expensive due to the complex nature of reservoir conditions. In contrast, methods that utilize historical production data to identify inter-well connectivity offer faster and more cost-effective alternatives. However, when faced with incomplete dynamic data—such as long-term shut-ins and data gaps—these methods may yield substantial errors in correlation results. To address this issue, we have developed an unsupervised machine learning algorithm that integrates sparse inverse covariance estimation with affinity propagation clustering to map and analyze dynamic oil field data. This methodology enables the extraction of inter-well topological structures, facilitating the automatic clustering of producers and the quantitative identification of connectivity between injectors and producers. To mitigate errors associated with sparse production data, our approach employs sparse inverse covariance estimation for preprocessing the production performance data of the wells. This preprocessing step enhances the robustness and accuracy of subsequent clustering and connectivity analyses. The algorithm’s stability and reliability were rigorously evaluated using long-term tracer test results from a test block in an actual reservoir, covering a span of over a decade. The results of the algorithm were compared with those of the tracer test to evaluate its accuracy, precision rate, recall rate, and correlation. The clustering results indicate that wells with similar characteristics and production systems are automatically grouped into distinct clusters, reflecting the underlying geological understanding. The algorithm successfully divided the test block into four macro-regions, consistent with geological interpretations. Furthermore, the algorithm effectively identified the inter-well connectivity between injectors and producers, with connectivity magnitudes aligning closely with actual tracer test data. Overall, the algorithm achieved a precision rate of 79.17%, a recall rate of 90.48%, and an accuracy of 91.07%. This congruence validates the algorithm’s effectiveness in the quantitative analysis of inter-well connectivity and demonstrates significant potential for enhancing the accuracy and efficiency of inter-well connectivity identification.

Too-Short-Arc Association and Clustering Method for Space-Based Optical Observations

Space-based optical observation is assuming an increasingly significant role in space surveillance and tracking. However, it often yields tracks that are too short, which are known as too short arcs (TSAs). A single TSA lacks sufficient information for reliable initial orbit determination of a resident space object. Therefore, it is typically necessary to perform tracklet association during catalog building and maintenance. This study presents a novel tracklet association method that improves upon existing techniques by combining admissible regions and nonlinear orbital uncertainty propagation. Proceeding by constructing TSA association matrices and TSA clustering matrices, the bond energy algorithm, traditionally employed in the design of distributed database systems, is applied to cluster TSAs based on the association results between two arbitrary TSAs. Furthermore, the proposed splitting algorithm reduces the computational load associated with clustering multiple tracklets and effectively handles erroneous association results. To assess the effectiveness of this approach, it was tested in a space-based optical-survey scenario, accounting for practical observation uncertainties. The simulation results demonstrate that the method achieves a high level of accuracy in both TSA association and clustering.

ScDRMAE: integrating masked autoencoder with residual attention networks to leverage omics feature dependencies for accurate cell clustering.

Cell clustering is foundational for analyzing the heterogeneity of biological tissues using single-cell sequencing data. With the maturation of single-cell multi-omics sequencing technologies, we can integrate multiple omics data to perform cell clustering, thereby overcoming the limitations of insufficient information from single omics data. Existing methods for cell clustering often only consider the differences in data patterns during the analysis of multi-omics data, but the dependencies between omics features of different cell types also significantly influence cell clustering. Moreover, the high dropout rates in scRNA-seq and scATAC-seq data can impact the performance of cell clustering. We propose a cell clustering model based on a masked autoencoder, scDRMAE. Utilizing a masking mechanism, scDRMAE effectively learns the relationships between different features and imputes false zeros caused by dropout events. To differentiate the importance of various omics data in cell clustering, we dynamically adjust the weights of different omics data through an attention mechanism. Finally, we use the K-means algorithm for cluster analysis of the fused multi-omics data. On commonly used sets of 15 multi-omics datasets, our method demonstrates superior cell clustering performance on multiple metrics compared to other computational methods. In addition, when datasets exhibit varying degrees of dropout noise, our method shows better performance and stronger stability on multiple metrics compared to other methods. Moreover, by analyzing the cell clusters classified by scDRMAE, we identified several biologically significant biomarkers that have been validated, further confirming the effectiveness of scDRMAE in cell clustering from a biological perspective.