Clustering Method Research Articles

Clustering First-order reversal curve diagram using the Gaussian mixture model and the Davies Bouldin index

Abstract In our energy-consuming society, understanding magnetization reversal in permanent magnets is crucial for improving energy conversion efficiency between electric energy and mechanical energy. First-order reversal curves (FORCs) have enabled qualitative studies of reversal mechanisms, however, further understanding based on quantitative analysis is still difficult due to the complexity of FORC diagrams. We introduce a machine-learning based approach combining the Gaussian mixture model and the Davies-Bouldin index to separate characteristic features in FORC diagrams of Nd-Fe-B magnets. The clustering method is evaluated using several FORC diagrams obtained at different temperatures, and we also demonstrated hysteresis loop reconstruction using the clustered FORC diagrams.

Scientometrics Analysis of Global Researches on Anemia

Background: Anemia is a condition in which the number of red blood cells or the hemoglobin concentration within them is lower than normal. This study aims to show the intellectual structure of knowledge regarding anemia and gives a comprehensive and up-to-date image of research in this area. Materials and Methods: This is a descriptive-analytical study with a scientometric approach. The PubMed database was searched for research publications indexed under "anemia" including 8484 records between 2011 and 2020. Data were analyzed using Co-word analysis, clustering methods, and strategic diagrams with the help of SPSS and Ucinet 6 software. Results: The keyword “Anemia Sickle Cell” and two pairs of frequently used keywords, namely “Anemia, Iron *Iron” were the most frequent in the research area. The results shaped the concepts of anemia in 9 clusters. The clusters "Hydroxyurea and sickle cell anemia", "Fetus transfusion", "Management of Thalassemia Major", "Hemolytic Uremic Syndrome", "Management and Control of Anemia", "Chronic Kidney Failure and Anemia", "Hematopoietic Stem Cell Transplantation" are topics that may be emerging or disappearing. The “Thalassemia and blood transfusion" are immature clusters. Conclusion: This study uses co-word networks that indicate important links between keywords of the research areas. Most research approaches are in the therapeutic aspects. Despite the importance of the effect of anemia on all levels of society, including economics, education, and other types of anemia, as well as its impact on learning and mental disorders, these subjects have not been given sufficient consideration.

Sentence Embeddings and Semantic Entity Extraction for Identification of Topics of Short Fact-Checked Claims

The objective of this research was to design a method to assign topics to claims debunked by fact-checking agencies. During the fact-checking process, access to more structured knowledge is necessary; therefore, we aim to describe topics with semantic vocabulary. Classification of topics should go beyond simple connotations like instance-class and rather reflect broader phenomena that are recognized by fact checkers. The assignment of semantic entities is also crucial for the automatic verification of facts using the underlying knowledge graphs. Our method is based on sentence embeddings, various clustering methods (HDBSCAN, UMAP, K-means), semantic entity matching, and terms importance assessment based on TF-IDF. We represent our topics in semantic space using Wikidata Q-ids, DBpedia, Wikipedia topics, YAGO, and other relevant ontologies. Such an approach based on semantic entities also supports hierarchical navigation within topics. For evaluation, we compare topic modeling results with claims already tagged by fact checkers. The work presented in this paper is useful for researchers and practitioners interested in semantic topic modeling of fake news narratives.

Exploring electronic resonances in pyridine: Insights from orbital stabilization techniques.

Electron attachment to pyridine results in electronic resonances, metastable states that can decay through electronic or nuclear degrees of freedom. This study uses orbital stabilization techniques combined with bound electronic structure methods, based on equation of motion coupled cluster or multi-reference methods, to calculate positions and widths of electronic resonances in pyridine that exist below 10eV. We report four 2B1 and four 2A2 resonances, including one 2B1 not previously reported experimentally and two 2A2 resonances not reported at all in the literature. The two lower energy resonances are one-particle shape resonances, while the remaining are mixed or primarily core-excited resonances. Multi-reference perturbation theory provides the best description of these resonances, especially when their character is mixed. We describe the character of these resonances qualitatively and calculate Dyson orbitals, which provide information about their decay channels.

Tensorized diversity and consistency with Laplacian manifold for multi-view clustering

The advantage of multi-view clustering lies in its ability to leverage the diversity and consistency among multiple views to better capture the intrinsic structure of the data. However, existing multi-view methods treat diversity and consistency as a set of opposing attributes, overlooking their inherent connections. Meanwhile, the complete information across multiple views is not fully utilized. To address these issues, this paper proposes the tensorized diversity and consistency with Laplacian manifold for multi-view clustering method (TDCLM). Specifically, starting from the self-expressive property of the original data, we obtain the diversity graphs and the consistency graph, and for the first time, we combined Laplacian manifold constraints to strengthen the relationship between diversity and consistency while jointly optimizing the diversity graphs and the consistency graph. Additionally, we innovatively combine the diversity graphs and the consistency graph into a tensor and subject it to the constraint of tensor nuclear norm. By doing so, we not only obtain the complete information between multiple views but also enable the mutual learning and mutual enhancement of the diversity graphs and the consistency graph. Finally, by adopting the augmented Lagrange multiplier method, we integrate the two steps into a comprehensive framework. The TDCLM shows a performance enhancement of up to 25.85%, with experimental results across diverse datasets demonstrating that the TDCLM algorithm surpasses the state-of-the-art algorithms. In other words, these experimental results validate the importance of obtaining complete information from multiple views and effectively leveraging the diversity and consistency inherent in this complete information. The code is publicly available at https://github.com/TongWuahpu/TDCLM.

Computing the Commonalities of Clusters in Resource Description Framework: Computational Aspects

Clustering is a very common means of analysis of the data present in large datasets, with the aims of understanding and summarizing the data and discovering similarities, among other goals. However, despite the present success of the use of subsymbolic methods for data clustering, a description of the obtained clusters cannot rely on the intricacies of the subsymbolic processing. For clusters of data expressed in a Resource Description Framework (RDF), we extend and implement an optimized, previously proposed, logic-based methodology that computes an structure—called a Common Subsumer—describing the commonalities among all resources. We tested our implementation with two open, and very different, datasets: one devoted to public procurement, and the other devoted to drugs in pharmacology. For both datasets, we were able to provide reasonably concise and readable descriptions of clusters with up to 1800 resources. Our analysis shows the viability of our methodology and computation, and paves the way for general cluster explanations to be provided to lay users.

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.

Towards determining the presence of barren plateaus in some chemically inspired variational quantum algorithms

In quantum chemistry, the variational quantum eigensolver (VQE) is a promising algorithm for molecular simulations on near-term quantum computers. However, VQEs using hardware-efficient circuits face scaling challenges due to the barren plateau problem. This raises the question of whether chemically inspired circuits from unitary coupled cluster (UCC) methods can avoid this issue. Here we provide theoretical evidence indicating they may not. By examining alternated dUCC ansätzes and relaxed Trotterized UCC ansätzes, we find that in the infinite depth limit, a separation occurs between particle-hole one- and two-body unitary operators. While one-body terms yield a polynomially concentrated energy landscape, adding two-body terms leads to exponential concentration. Numerical simulations support these findings, suggesting that popular 1-step Trotterized unitary coupled-cluster with singles and doubles (UCCSD) ansätze may not scale. Our results emphasize the link between trainability and circuit expressiveness, raising doubts about VQEs’ ability to surpass classical methods.

Investigating the relationship between internet addiction, domestic violence, and emotional divorce among married women in Tehran

BackgroundIdentifying the factors affecting emotional divorce due to the negative consequences of this social problem in women can play an important role in preventing its occurrence. Therefore, the present study aimed to investigate the relationship between internet addiction, domestic violence, and emotional divorce among married women in Tehran.MethodsA cross-sectional survey was conducted among married women in Tehran, Iran. Sampling was conducted first through a multi-stage cluster method and then continued using a random method. Tehran was divided into five parts: central, northern, southern, eastern, and western. The sample size was divided by 5, resulting in 80 participants selected from each part. In each part, two regions were selected, and in each region, 40 houses were randomly selected. Researchers visited the selected houses, and the questionnaires were filled out at home by the women. Tools such as Gottman’s Emotional Divorce Test, Young’s Internet Addiction Test, and the Domestic Violence against Women Survey were used to collect data. Descriptive and inferential statistics were used to analyze the data in SPSS (version 24) software.ResultsA total of 400 women participated in the study. 46% of all married women suffered emotional divorce. The analysis of these tests showed that are positive and significant relationships between emotional divorce and internet addiction (r = 0.626, p < 0.0001), emotional divorce and domestic violence (r = 0.632, p < 0.0001), and internet addiction and domestic violence (r = 0.455, p < 0.0001). It was also found that there are significant negative relationships between emotional divorce and both the level of education (r=-0.115, p = 0.021) and employment status (r=-0.512, p < 0.0001). Additionally, there are significant negative relationships between internet addiction and age (r=-0.285, p < 0.0001), duration of marriage (r=-0.237, p < 0.0001) and employment status (r=-0.694, p < 0.0001). Furthermore, it was revealed that there are significant negative relationship between domestic violence and age (r=-0.104, p = 0.037) and level of education (r=-0.156, p = 0.002).ConclusionAlmost half of the married women in this study experienced emotional divorce, internet addiction, and domestic violence. These factors have a direct relationship with emotional divorce. Future studies are needed to understand how internet addiction might be associated with separation, emotional divorce and family problems.

Optimization of reduction gear in anchor winch based on modal analysis

In order to achieve more scientific design of the reduction gear and reduce material waste, pre-stressed modal analysis method was combined with multi-objective optimization algorithm to optimize the structure of the reduction gear basic body. The model was simplified and parameterized and the maximum stress and equivalent stiffness under different parameter size combinations were obtained through finite element analysis. Separately, genetic clustering method, neural network method, and Kriging method were used to construct the response surface function. Through error verification and comparison, it was found that the Kriging method was more suitable for the gear model. In the design of variable extremum search, multi-objective genetic algorithm and sequential quadratic programming were compared and analyzed. The results show that the mass of the gear can be reduced by 39.9 %, while the maximum stress remains unchanged, the equivalent stiffness is not reduced, and a good optimization design effect is achieved.

Discovering Electric Vehicle Charging Locations Based on Clustering Techniques Applied to Vehicular Mobility Datasets

With the proliferation of vehicular mobility traces because of inexpensive on-board sensors and smartphones, utilizing them to further understand road movements have become easily accessible. These huge numbers of vehicular traces can be utilized to determine where to enhance road infrastructures such as the deployment of electric vehicle (EV) charging stations. As more EVs are plying today’s roads, the driving anxiety is minimized with the presence of sufficient charging stations. By correctly extracting the various transportation parameters from a given dataset, one can design an adequate and adaptive EV charging network that can provide comfort and convenience for the movement of people and goods from one point to another. In this study, we determined the possible EV charging station locations based on an urban city’s vehicular capacity distribution obtained from taxi and ride-hailing mobility GPS traces. To achieve this, we first transformed the dynamic vehicular environment based on vehicular capacity into its equivalent urban single snapshot. We then obtained the various traffic zone distributions by initially utilizing k-means clustering to allow flexibility in the total number of wanted traffic zones in each dataset. In each traffic zone, iterative clustering techniques employing Density-based Spatial Clustering of Applications with Noise (DBSCAN) or clustering by fast search and find of density peaks (CFS) revealed various area separation where EV chargers were needed. Finally, to find the exact location of the EV charging station, we last ran k-means to locate centroids, depending on the constraint on how many EV chargers were needed. Extensive simulations revealed the strengths and weaknesses of the clustering methods when applied to our datasets. We utilized the silhouette and Calinski–Harabasz indices to measure the validity of cluster formations. We also measured the inter-station distances to understand the closeness of the locations of EV chargers. Our study shows how CFS + k-means clustering techniques are able to pinpoint EV charger locations. However, when utilizing DBSCAN initially, the results did not present any notable outcome.

Discovering Hidden Patterns: Applying Topic Modeling in Qualitative Research

In qualitative studies, researchers must devote a significant amount of time and effort to extracting meaningful themes from huge sets of texts and examining the links between themes, which are frequently done manually. The availability of natural language models has enabled the application of a wide range of techniques for automatically detecting hierarchy, linkages, and latent themes in texts. This paper aims to investigate the coherence of the topics acquired from the analysis with the predefined themes, the hierarchy between the topics, the similarity between the topics and the proximity-distance between the topics by means of the topic model based on BERTopic using unstructured qualitative data. The qualitative data for this study was gathered from 106 students engaged in a university-run pedagogical formation certificate program. In BERTopic procedure, paraphrase-multilingual-MiniLM-L12-v2 model was used as sentence transformer model, UMAP was used as dimension reduction method and HDBSCAN algorithm was used as clustering method. It is found that BERTopic successfully identified six topics corresponding to the six predicted themes in unstructured texts. Moreover 74% of the texts containing some themes could be classified accurately. The algorithm was also able to successfully identify which topics were similar and which topics differed significantly from the others. It was concluded that BERTopic is a procedure that can identify themes that researchers do not notice depending on the density of the data in qualitative data analysis and has the potential to enable qualitative research to reach more detailed findings.

Forming Dataset of The Undergraduate Thesis using Simple Clustering Methods

Each university collects many undergraduate theses data but has yet to process it to make it easier forstudents to find references as desired. This study aims to classify and compare the grouping ofdocuments using expert and simple clustering methods. Experts have done ground truth using ORBoolean Retrieval and keyword generation. The best cluster was discovered by the experiments usingthe K-Means, K-Medoids, and DBSCAN clustering methods and using Euclidean, Manhattan, CityBlock, and Cosine Similarity metrics. The cluster with the best Silhouette Score compared to theaccuracy of the categorization of each document. The K-Means clustering method and the CosineSimilarity metric gave the best results with a Silhouette Score value of 0.105534. The comparisonbetween ground truth and the best cluster results shows an accuracy of 33.42%. The result shows thatthe simple clustering method cannot handle data with Negative Skewness and Leptokurtic Kurtosis.

ОСОБЛИВОСТІ ВИБОРУ СТРАТЕГІЙ ПОЗИЦІОНУВАННЯ МАЛИХ ФЕРМЕРСЬКИХ ГОСПОДАРСТВ-ВИРОБНИКІВ ЯБЛУК

For a long time, small farms were left out of marketing strategies, focusing on cultivation and traditional schemes for the sale of their products. Marketing entered agricultural industries later than industrial and consumer markets, which creates gaps in the use of marketing tools. Modern challenges actualize the mechanisms of selection, development and implementation of positioning strategies for small farms. The purpose of the article is to research and analyze the features of the market and marketing environment in which small apple farms operate and, accordingly, research the impact of the identified features on the choice of positioning strategies of these enterprises. The object of the study is positioning strategies of small farms producing apples. Methods used in the research: desk analysis, clustering method, target audience segmentation method, methods of analysis, synthesis and logical generalization, method of generalization of national experience, etc. The hypothesis of the study was the assumption that relationship marketing at the level of the B2B segment imposes its specificity on all areas of marketing activity of small farms producing apples, including on the processes of their positioning. Main Content. Classical strategies are not always applicable in the field of agricultural production, where success often depends on the ability to establish connections and cooperation with intermediaries, distribution partners, service organizations, local authorities, the labor market, and others. This necessitates the more active use of positioning strategies by agricultural producers. Apple farming in Ukraine is a strong sector of horticulture and agricultural production as a whole. It ensures the country's food security, creates significant export flows, supplies raw materials to the food industry, is a major consumer of machinery and chemicals, and provides substantial employment. Currently, a significant number of agricultural enterprises in the Chernivtsi region are engaged in apple farming. It has been established that all apple producers in this region can be categorized as small farms. It is noted that widely accepted positioning strategies need to be adapted to meet the needs of small-scale agricultural producers. The article emphasizes that among all possible positioning strategies, only a limited number are suitable for them—those focusing on the quality of the product, the "quality" of the supplier, the "quality" of the partner, and the "quality" of the employer. Given that positioning strategies require targeting a specific audience, the author identifies the following target audiences for small farms: 1) Wholesale intermediaries and buyers, 2) Local competitors, 3) Permanent and seasonal workers, 4) The local community. Originality and Practical Significance of the Study. The research analyzes practical experience and takes into account the current realities of the agricultural sector to develop effective recommendations for its application in the strategic planning of the mentioned enterprises. Conclusions and prospects for further research. The research on the characteristics of the highly competitive apple agrimarket justifies the need for small farms to develop a comprehensive positioning strategy, which includes the following components: positioning the farm as a reliable partner for wholesalers and intermediaries; positioning the farm as a trustworthy partner for cooperation and collaboration; positioning the farm as a dependable employer; and positioning the farm as a producer and supplier of high-quality products.

Wood trade regions of Russia: clustering approach development

Network analysis of the Russian timber industry was based on the information about timber transactions between Russian companies in 2020. The data were collected from the Unified State Automated Information System Accounting Timber and Transactions with It. The method of graph clustering was developed. The first step is clustering of the hole graph by Leiden algorithm. The second step is drawing of each Leiden created cluster in abstract space using Fruchterman–Reingold layout and extraction of clusters from layout using meanshift algorithm. Such approach helped us to divide many Leiden-created clusters into components. This algorithm has a problem of resolution limit. It tends to combine some medium and small clusters into large ones. Also, a special method was developed to limit the uncertainty of clustering – both Leiden algorithm and Fruchterman–Reingold layout are non-deterministic algorithms. Wood trade clusters vary in size and content. They are rather autonomous. The average share of intra-cluster trade is 89%. The modularity score of our partition is 0.863. Each cluster has been described using open sources from the Internet. The content of the clusters does not contradict the logic of the wood industry processes. The clusters create rather compact areas on the map – wood trade regions. Their borders often overlap with existing administrative boundaries. Five types of network structure of clusters were defined: vertical monocentric, vertical polycentric, horizontal, dendritic, simple. There are three factors of clustering in wood industry of Russia: production chains (4 types – pulp, plywood, lumber, chipboard chains), demand chains (redistribution of raw material between manufacturers within the cluster, accumulation of wood volumes for some enterprises outside the cluster, wood trade between traders within the cluster), common holder. We have shown that existing approaches to cluster detection based on location quotients often misrepresent economic networks.

A Radial Distribution Function Based Recognition Algorithm of Point Defects in Large-Scale β-Ga2O3 Systems.

The atomic configurations and concentrations of intrinsic defects profoundly influence the electrical and optical properties of the semiconductor materials. This influence is particularly significant in the case of β-Ga2O3, which is a highly promising ultrawide bandgap semiconductor characterized by highly complex intrinsic defect configurations. Despite its importance, there is a notable absence of an accurate method to recognize these defects in large-scale atomistic computational modeling. We design an effective algorithm for the explicit identification of various intrinsic point defects in the β-Ga2O3 lattice, which constitutes the integration of the particle swarm optimization (PSO) and K-means clustering (K-MC) methods. Our algorithm attains the recognition accuracy exceeding 95%. Finally, the algorithm is applied to dynamic simulations, where the feasibility of dynamic real-time detection is explored.

A Comprehensive Survey on Deep Clustering: Taxonomy, Challenges, and Future Directions

Clustering is a fundamental machine learning task which aims at assigning instances into groups so that similar samples belong to the same cluster while dissimilar samples belong to different clusters. Shallow clustering methods usually assume that data are collected and expressed as feature vectors within which clustering is performed. However, clustering high-dimensional data, such as images, texts, videos, and graphs, poses significant challenges for clustering tasks, such as indiscriminate representation and intricate relationships among instances. Over the past decades, deep learning has achieved remarkable success in effective representation learning and modeling complex relationships. Motivated by these advancements, Deep Clustering seeks to improve clustering outcomes through deep learning techniques, garnering considerable interest from both academia and industry. Despite many contributions to this vibrant area of research, the lack of systematic analysis and a comprehensive taxonomy has hindered progress in this field. In this survey, we first explore how deep learning can be integrated into deep clustering and identify two fundamental components: the representation learning module and the clustering module. Then we summarize and analyze the representative design of these two modules. Furthermore, we introduce a novel taxonomy of deep clustering based on how these two modules interact, specifically through multistage, generative, iterative, and simultaneous approaches. In addition, we present well-known benchmark datasets, evaluation metrics, and open-source tools to clearly demonstrate different experimental approaches. Finally, we examine the practical applications of deep clustering and propose challenging areas for future research.

Experimental and statistical methods for microrheological characterization of heterogeneity in human respiratory mucus mimics of health and disease progression

Human respiratory mucus (HRM) is extremely soft, compelling passive microrheology for linear viscoelastic characterization. We focus this study on the use of passive microrheology to characterize HRM heterogeneity, a phenomenon in normal HRM that becomes extreme during cystic fibrosis (CF) disease. Specifically, a fraction of the mucin polymers comprising HRM phase-separate into insoluble structures, called flakes, dispersed in mucin-depleted solution. We first reconstitute HRM samples to the MUC5B:MUC5AC mucin ratios consistent with normal and CF clinical samples, which we show recapitulate progressive flake formation and heterogeneity. We then employ passive particle tracking with 200 nm and 1 μm diameter beads in each reconstituted sample. To robustly analyze the tracking data, we introduce statistical denoising methods for low signal-to-noise tracking data within flakes, tested and verified using model-generated synthetic data. These statistical methods provide a fractional Brownian motion classifier of all successfully denoised, tracked beads in flakes and the dilute solution. From the ensemble of classifier data, per bead diameter and mucus sample, we then employ clustering methods to learn and infer multiple levels of heterogeneity: (i) tracked bead data within vs. outside flakes and (ii) within-flake data buried within or distinguishable from the experimental noise floor. Simulated data consistent with experimental data (within and outside flakes) are used to explore form(s) of the generalized Stokes–Einstein relation (GSER) that recover the dynamic moduli of homogeneous and heterogeneous truth sets of purely flakelike, dilute solution, and mixture samples. The appropriate form of GSER is applied to experimental data to show (i) flakes are heterogeneous with gel and sol domains; (ii) dilute solutions are heterogeneous with only sol domains; and (iii) flake and dilute solution properties vary with probe diameter.

Investigating TCR-pMHC interactions for TCRs without identified epitopes by constructing a computational pipeline

The molecular mechanisms underlying epitope recognition by T cell receptors (TCRs) are critical for activating T cell immune responses and rationally designing TCR-based therapeutics. Single-cell sequencing techniques vastly boost the accumulation of TCR sequences, while the limitation of available TCR-pMHC structures hampers further investigations. In this study, we proposed a computational pipeline that incorporates structural information and single-cell sequencing data to investigate the epitope-recognition mechanisms for TCRs without identified epitopes. By antigen specificity clustering, we mapped the epitope sequences between epitope-known and epitope-unknown TCRs from COVID-19 patients. One reported SARS-CoV-2 epitope, NQKLIANQF (S919–927), was identified for a TCR expressed by 614 T cells (TCR-614). Epitope screening also identified a potential cross-reactive epitope, KLKTLVATA (NSP31790–1798), for a TCR expressed by 204 T cells (TCR-204). By molecular dynamics (MD) simulations, we revealed the detailed epitope-recognition mechanisms for both TCRs. The structural motifs responsible for epitope recognition revealed by the MD simulations are consistent with the sequential features recognized by the sequence-based clustering method. We hope that this strategy could facilitate the discovery and optimization of TCR-based therapeutics.

DGSIST: Clustering spatial transcriptome data based on deep graph structure Infomax

Although spatial transcriptomics data provide valuable insights into gene expression profiles and the spatial structure of tissues, most studies rely solely on gene expression information, underutilizing the spatial data. To fully leverage the potential of spatial transcriptomics and graph neural networks, the DGSI (Deep Graph Structure Infomax) model is proposed. This innovative graph data processing model uses graph convolutional neural networks and employs an unsupervised learning approach. It maximizes the mutual information between graph-level and node-level representations, emphasizing flexible sampling and aggregation of nodes and their neighbors. This effectively captures and incorporates local information from nodes into the overall graph structure. Additionally, this paper developed the DGSIST framework, an unsupervised cell clustering method that integrates the DGSI model, SVD dimensionality reduction algorithm, and k-means++ clustering algorithm. This aims to identify cell types accurately. DGSIST fully uses spatial transcriptomics data and outperforms existing methods in accuracy. Demonstrations of DGSIST’s capability across various tissue types and technological platforms have shown its effectiveness in accurately identifying spatial domains in multiple tissue sections. Compared to other spatial clustering methods, DGSIST excels in cell clustering and effectively eliminates batch effects without needing batch correction. DGSIST excels in spatial clustering analysis, spatial variation identification, and differential gene expression detection and directly applies to graph analysis tasks, such as node classification, link prediction, or graph clustering. Anticipation lies in the contribution of the DGSIST framework to a deeper understanding of the spatial organizational structures of diseases such as cancer.