Mixture Model-based Clustering Research Articles

A Gibbs Posterior Framework for Fair Clustering.

The rise of machine learning-driven decision-making has sparked a growing emphasis on algorithmic fairness. Within the realm of clustering, the notion of balance is utilized as a criterion for attaining fairness, which characterizes a clustering mechanism as fair when the resulting clusters maintain a consistent proportion of observations representing individuals from distinct groups delineated by protected attributes. Building on this idea, the literature has rapidly incorporated a myriad of extensions, devising fair versions of the existing frequentist clustering algorithms, e.g., k-means, k-medioids, etc., that aim at minimizing specific loss functions. These approaches lack uncertainty quantification associated with the optimal clustering configuration and only provide clustering boundaries without quantifying the probabilities associated with each observation belonging to the different clusters. In this article, we intend to offer a novel probabilistic formulation of the fair clustering problem that facilitates valid uncertainty quantification even under mild model misspecifications, without incurring substantial computational overhead. Mixture model-based fair clustering frameworks facilitate automatic uncertainty quantification, but tend to showcase brittleness under model misspecification and involve significant computational challenges. To circumnavigate such issues, we propose a generalized Bayesian fair clustering framework that inherently enjoys decision-theoretic interpretation. Moreover, we devise efficient computational algorithms that crucially leverage techniques from the existing literature on optimal transport and clustering based on loss functions. The gain from the proposed technology is showcased via numerical experiments and real data examples.

A Modified Mixture Model-Based Clustering Algorithm for Resolving the Problem of Mixed Pixels Available in Satellite Imagery

A Modified Mixture Model-Based Clustering Algorithm for Resolving the Problem of Mixed Pixels Available in Satellite Imagery

Ni2+ and Cu2+ complexes of N-(2,6-dichlorophenyl)-N-mesityl formamidine dithiocarbamate structural and functional properties as CYP3A4 potential substrates

Metal compounds continued to attract diverse applications due to their malleability in several capacities. In this study, we present our findings on the crystal structures and functional properties of Ni2+ and Cu2+ complexes of N'-(2,6-dichlorophenyl)-N-mesitylformamidine dithiocarbamate (L) comprising [Ni-(L)2] (1) and [Cu-(L)2] (2) with a four-coordinate metal center. We established the two complex structures through 1H and 13C nuclear magnetic resonance (NMR), elemental, and single-crystal X-ray analysis. The analyses showed that the two complexes are isomorphous, having P21/c as a space group and a unit-cell similarity index (π) of 0.002. The two complexes conform to a distorted square planar geometry around the metal centers. The calculated and experimental data, including bond lengths, angles, and NMR values, are similar. Hirshfeld surface analysis revealed the variational contribution of the different types of intermolecular contacts driven by the crystal lattice of the two solvated complexes. Our knowledge of the potential biological implication of these structures enabled us to probe the compounds as prospective CYP3A4 inhibitors. This approach mimics current trends in pharmaceutical design and biomedicine by incorporating potentially active molecules into various media to predict their biological efficacies. The simulations show appreciable binding of compounds 1 and 2 to CYP3A4 with average interaction energies of –97 and –87 kcal/mol, respectively. The protein attains at least five conformational states in the three studied models using a Gaussian Mixture Model-based clustering and free energy prediction. Electric field analysis shows the crucial residues to substrate binding at the active site, enabling CYP3A4 structure to function prediction. The predicted inhibition with these Ni2+ and Cu2+ complexes indicates that CYP3A4 overexpression in a diseased state like cancer would reduce, thereby increasing the chemotherapeutic compounds' shelf-lives for adsorption. This multidimensional study addresses various aspects of molecular metal electronics, including their application as substrate-mimicking inhibitors. The outcome would enable further research on bio-metal compounds of critical potential.

Identifying urban form typologies in Seoul using a new Gaussian mixture model-based clustering framework

Seoul, the capital city of South Korea, has diverse urban forms developed through its complex history. Previous studies show limitations of strong subjectivity and difficulty in scalability in identifying typical Seoul urban forms with expert knowledge. Data-driven approach offers an opportunity to address those challenges, but previous studies often focused on direct applications of clustering algorithms to a given area with diverse methods and workflows, lacking a systematic framework. This study addressed these issues by developing a new form clustering framework to systematically identify form typologies at a large scale and demonstrated its application in Seoul. With a 500 m × 500 m grid as the basic spatial unit and twelve urban form attributes as learning features, 14 clusters were identified using the Gaussian mixture model. These clusters were further translated into form typologies following a semantic typology naming system, with representative form samples identified. The resulting typologies were then verified and validated through comparisons with previous studies. Their relationships with zoning classes were also examined, emphasizing their role in urban planning and design. Results suggest this new framework is an effective and promising way to identify urban form typologies in complex urban environments to better support urban planning and management.

House Prices Segmentation Using Gaussian Mixture Model-Based Clustering

House is a place for humans to live and a main necessity for humans. For years, the need for houses is increasing and varied so that it affects the selling price of the house. Therefore, more research is needed to learn about the selling price of houses. This research is only focusing on house price segmentation in DKI Jakarta using the Gaussian Mixture Model-Based Clustering Method with the Expectation-Maximization algorithm. The goal of this research is to make a house price segmentation model so that we can obtain useful information for the potential buyer. Clustering with GMM utilize the log-likelihood function to optimize the GMM parameters. The result of this research is houses in DKI Jakarta can be segmented into 3 different clusters. The first cluster is for the low-profile houses. The second cluster is for the mid-profile houses. The third cluster is for the high-profile houses. The silhouette score that was produced by the clustering method is 0.60866 meaning that this score is quite good because it’s close to a value of 1.

Neural mixture models with expectation-maximization for end-to-end deep clustering

Any clustering algorithm must synchronously learn to model the clusters and allocate data to those clusters in the absence of labels. Mixture model-based methods model clusters with pre-defined statistical distributions and allocate data to those clusters based on the cluster likelihoods. They iteratively refine those distribution parameters and member assignments following the Expectation-Maximization (EM) algorithm. However, the cluster representability of such hand-designed distributions that employ a limited amount of parameters is not adequate for most real-world clustering tasks. In this paper, we realize mixture model-based clustering with a neural network where the final layer neurons, with the aid of an additional transformation, approximate cluster distribution outputs. The network parameters pose as the parameters of those distributions. The result is an elegant, much-generalized representation of clusters than a restricted mixture of hand-designed distributions. We train the network end-to-end via batch-wise EM iterations where the forward pass acts as the E-step and the backward pass acts as the M-step. In image clustering, the mixture-based EM objective can be used as the clustering objective along with existing representation learning methods. In particular, we show that when mixture-EM optimization is fused with consistency optimization, it improves the sole consistency optimization performance in clustering. Our trained networks outperform single-stage deep clustering methods that still depend on k-means, with unsupervised classification accuracy of 63.8% in STL10, 58% in CIFAR10, 25.9% in CIFAR100, and 98.9% in MNIST.

A hybrid steady-state evolutionary algorithm using random swaps for Gaussian model-based clustering

A new hybrid evolutionary algorithm (EA) for Gaussian mixture model-based clustering is proposed. The EA is a steady-state method that, in each generation, selects two individuals from a population, creates two offspring using either mutation or crossover, and fine-tunes the offspring using the expectation maximization (EM) algorithm. The offspring compete with their parents for survival into the next generation. The approach proposed uses a random swap, which replaces a component mean with a randomly chosen feature vector as a mutation operator. In the crossover operator, a random component is copied from the source mixture into a destination mixture. Copying in crossover favors components of the source mixture located away from the components of the destination mixture. In computational experiments, the approach was compared to a multiple restarts EM algorithm, a random swap EM method, and a state-of-the-art hybrid evolutionary algorithm for Gaussian mixture model learning on one real and 29 synthetic datasets. The results indicate that, given the same computational budget, the proposed method usually learns mixtures with higher log-likelihoods than other benchmarks competing algorithms. The partitions of data obtained by the method correspond best to the original divisions of datasets into classes.

Two Dimensional Clustering of Swift/BAT and Fermi/GBM Gamma-ray Bursts

Studies of Gamma-ray Burst (GRB) properties, such as duration and spectral hardness, have found evidence for additional classes beyond the short-hard (merger) and long-soft (collapsar) prototypes. Several clustering analyses of the duration-hardness plane identified a third, intermediate duration, class. In this work, Gaussian Mixture Model-based (GMM) clustering is applied to the Swift/BAT and Fermi/GBM samples of GRBs. The results obtained by the hierarchical combination of Gaussian components (or clusters) based on an entropy criterion are presented. This method counteracts possible overfitting arising from the application of Gaussian models to non-Gaussian underlying data. While the initial GMM clustering of the hardness-duration plane identifies three components (short/intermediate/long) for the Swift/BAT and Fermi/GBM samples, only two components (short/long) remain once the entropy criterion is applied. The analysis presented here suggests that the intermediate duration class may be the result of overfitting, rather than evidence of a distinct underlying population.

Discovering Thematically Coherent Biomedical Documents Using Contextualized Bidirectional Encoder Representations from Transformers-Based Clustering.

The increasing expansion of biomedical documents has increased the number of natural language textual resources related to the current applications. Meanwhile, there has been a great interest in extracting useful information from meaningful coherent groupings of textual content documents in the last decade. However, it is challenging to discover informative representations and define relevant articles from the rapidly growing biomedical literature due to the unsupervised nature of document clustering. Moreover, empirical investigations demonstrated that traditional text clustering methods produce unsatisfactory results in terms of non-contextualized vector space representations because that neglect the semantic relationship between biomedical texts. Recently, pre-trained language models have emerged as successful in a wide range of natural language processing applications. In this paper, we propose the Gaussian Mixture Model-based efficient clustering framework that incorporates substantially pre-trained (Bidirectional Encoder Representations from Transformers for Biomedical Text Mining) BioBERT domain-specific language representations to enhance the clustering accuracy. Our proposed framework consists of main three phases. First, classic text pre-processing techniques are used biomedical document data, which crawled from the PubMed repository. Second, representative vectors are extracted from a pre-trained BioBERT language model for biomedical text mining. Third, we employ the Gaussian Mixture Model as a clustering algorithm, which allows us to assign labels for each biomedical document. In order to prove the efficiency of our proposed model, we conducted a comprehensive experimental analysis utilizing several clustering algorithms while combining diverse embedding techniques. Consequently, the experimental results show that the proposed model outperforms the benchmark models by reaching performance measures of Fowlkes mallows score, silhouette coefficient, adjusted rand index, Davies-Bouldin score of 0.7817, 0.3765, 0.4478, 1.6849, respectively. We expect the outcomes of this study will assist domain specialists in comprehending thematically cohesive documents in the healthcare field.

Spatio-Temporal Dynamics of European Innovation—An Exploratory Approach via Multivariate Functional Data Cluster Analysis

We apply a functional data approach for mixture model-based multivariate innovation clustering to identify different regional innovation portfolios in Europe, considering patterns of specialization among innovation types. We combine patent registration data and other innovation and economic data across 225 regions, 13 years, and eight patent classes. The approach allows us to form several regional clusters according to their specific innovation types and captures spatio-temporal dynamics too subtle for most other clustering methods. Consistent with the literature on innovation systems, our analysis supports the value of regionalized clusters that can benefit from flexible policy support to strengthen regions as well as innovation in a systematic context, adding technology specificity as a new criterion to consider. The regional innovation cluster solutions for IPC classes for ‘fixed constructions’ and ‘mechanical engineering’ are highly comparable but relatively less comparable for ‘chemistry and metallurgy’. The clusters for innovations in ‘physics’ and ‘chemistry and metallurgy’ are similar; innovations in ‘electricity’ and ‘physics’ show similar temporal dynamics. For all other innovation types, the regional clustering is different. By taking regional profiles, strengths, and developments into account, options for improved efficiency of location-based regional innovation policy to promote tailored and efficient innovation-promoting programs can be derived.

Compositional analysis of the tonsil microbiota in relationship to Streptococcus suis disease in nursery pigs in Ontario

BackgroundThe tonsil of the soft palate in pigs is the colonization site of both commensal and pathogenic microbial agents. Streptococcus suis infections are a significant economic problem in the swine industry. The development of S. suis disease remains poorly understood. The purpose of this study was to identify whether the tonsillar microbiota profile in nursery pigs is altered with S. suis disease. Here, the dynamics of the tonsillar microbiota from 20 healthy pigs and 43 diseased pigs with S. suis clinical signs was characterized.ResultsBased on the presence or absence of S. suis in the systemic sites, diseased pigs were classified into confirmed (n = 20) or probable (n = 23) group, respectively. Microbiota composition was assessed using the V3-V4 hypervariable region of the 16S rRNA, and results were analyzed to identify the diversity of the tonsillar microbiota. The taxonomic composition of the tonsil microbiota proved to be highly diverse between individuals, and the results showed statistically significant microbial community structure among the diagnosis groups. The confirmed group had the lowest observed species richness while the probable group had higher phylogenetics diversity level compared to the healthy group. Un-weighted Unifrac also demonstrated that the probable group had a higher beta diversity than both the healthy and the confirmed group. A Dirichlet-multinomial mixture (DMM) model-based clustering method partitioned the tonsil microbiota into two distinct community types that did not correspond with disease status. However, there was an association between Streptococcus suis serotype 2 and DMM community type 1 (p = 0.03). ANCOM-BC identified 24 Streptococcus amplicon sequence variants (ASVs) that were differentially abundant between the DMM community types.ConclusionsThis study provides a comprehensive analysis of the structure and membership of the tonsil microbiota in nursery pigs and uncovers differences and similarities across varying S. suis disease status. While the overall abundance of Streptococcus was not different among the diagnosis groups, the unique profile of DMM community type 1 and the observed correlation with S. suis serotype 2 could provide insight into potential tonsillar microbiota involvement in S. suis disease.

MBMM: Moment Estimating Beta Mixture Model-based Clustering Algorithm for m6A Co-methylation Module Mining

Background: m6A methylation is a ubiquitous post-transcriptional modification that exists in mammals. MeRIP-seq technology makes the acquisition of m6A data in the whole transcriptome under different conditions realizable. The specific regulation of the enzyme will present comethylation module on m6A methylation level data. Thus, mining the co-methylation module from which can help to unveil the mechanism of m6A methylation modification and its mechanism in the occurrence and development of complex diseases such as cancer. Objective: To develop a clustering algorithm that can effectively realize the mining of m6A comethylation module. Methods: In this study, a novel beta mixture model-based clustering algorithm named MBMM was proposed, which is based on the EM framework and introduces the method of moment estimating in M-step for parameter estimation to tackle the high-dimensional small sample m6A data. Simulation research was employed to evaluate the clustering performance of the proposed algorithm, and by which the co-methylation module mining was done based on real data. Biological significance correlation analysis was employed to explore whether the clustering results are co-methylation modules. Results and Conclusion: Simulation research demonstrated that MBMM performed out than other clustering algorithms. In real data, seven co-methylation modules were found by MBMM. Six m6Arelated pathways specific analysis showed that six co-methylation modules were enriched in the pathway and were different. Five enzymes substrate-specific analysis revealed that seven comethylation modules expressed varying degrees of enrichment. Gene Ontology enrichment analysis indicated that these modules may be regulated by enzymes while having potential functional specificity.

Consensus clustering for case series identification and adverse event profiles in pharmacovigilance

ObjectiveTo describe and evaluate vigiGroup - a consensus clustering algorithm which can identify groups of individual case reports referring to similar suspected adverse drug reactions and describe associated adverse event profiles, accounting for co-reported adverse event terms. Materials and methodsConsensus clustering is achieved by grouping pairs of reports that are repeatedly placed together in the same clusters across a set of mixture model-based cluster analyses. The latter use empirical Bayes statistical shrinkage for improved performance. As baseline comparison, we considered a regular mixture model-based cluster analysis. Three randomly selected drugs in VigiBase, the World Health Organization's global database of Individual Case Safety Reports were analyzed: sumatriptan, ambroxol and tacrolimus. Clustering stability was assessed using the adjusted Rand index, ranging between −1 and +1, and clinical coherence was assessed through an intruder detection analysis. ResultsFor the three drugs considered, vigiGroup achieved stable and coherent results with adjusted Rand indices between +0.80 and +0.92, and intruder detection rates between 86% and 94%. Consensus clustering improved both stability and clinical coherence compared to mixture model-based clustering alone. Statistical shrinkage improved the stability of clusters compared to the baseline mixture model, as well as the cross-validated log-likelihood. ConclusionsThe proposed algorithm can achieve adequate stability and clinical coherence in clustering individual case reports, thereby enabling better identification of case series and associated adverse event profiles in pharmacovigilance. The use of empirical Bayes shrinkage and consensus clustering each led to meaningful improvements in performance.

Classifying patients with depressive and anxiety disorders according to symptom network structures: A Gaussian graphical mixture model-based clustering.

Patients with mental disorders often suffer from comorbidity. Transdiagnostic understandings of mental disorders are expected to provide more accurate and detailed descriptions of psychopathology and be helpful in developing efficient treatments. Although conventional clustering techniques, such as latent profile analysis, are useful for the taxonomy of psychopathology, they provide little implications for targeting specific symptoms in each cluster. To overcome these limitations, we introduced Gaussian graphical mixture model (GGMM)-based clustering, a method developed in mathematical statistics to integrate clustering and network statistical approaches. To illustrate the technical details and clinical utility of the analysis, we applied GGMM-based clustering to a Japanese sample of 1,521 patients (Mage = 42.42 years), who had diagnostic labels of major depressive disorder (MDD; n = 406), panic disorder (PD; n = 198), social anxiety disorder (SAD; n = 116), obsessive-compulsive disorder (OCD; n = 66), comorbid MDD and any anxiety disorder (n = 636), or comorbid anxiety disorders (n = 99). As a result, we identified the following four transdiagnostic clusters characterized by i) strong OCD and PD symptoms, and moderate MDD and SAD symptoms; ii) moderate MDD, PD, and SAD symptoms, and weak OCD symptoms; iii) weak symptoms of all four disorders; and iv) strong symptoms of all four disorders. Simultaneously, a covariance symptom network within each cluster was visualized. The discussion highlighted that the GGMM-based clusters help us generate clinical hypotheses for transdiagnostic clusters by enabling further investigations of each symptom network, such as the calculation of centrality indexes.

The development of a representative multidimensional transient duty cycle for in-service switcher locomotives

Switcher locomotives operate in railway yards where they shunt railcars and assemble trains. Shunting railcars requires frequent aggressive acceleration and deceleration events in order for the locomotive to push or pull railcars onto specific tracks. As a result, switcher locomotives rarely sustain tractive power for any significant period of time. Given that all switchers in North America rely on diesel-electric propulsion; the result is rapid and frequent transitions in engine power leading to a very low engine efficiency and high levels of emissions. Any attempt to quantify or remedy these issues will face a lack of a representative profile or test cycle. A locomotive duty cycle is a breakdown of time spent at each power level of the locomotive’s engine. A major drawback of current duty cycles is that they only account for steady power. Such cycles are not representative of real switcher locomotive operation. This paper presents a real-world transient duty cycle for switcher locomotives that accounts for the rapid power transitions and is argued to be more statistically representative of actual operations. The methodology adopted relies on real-time data collection, microtrip based trip segmentation, and a finite mixture model-based clustering algorithm. Measurements were collected on a EMD 16-645 GP9 locomotive. The duty cycle developed herein is representative of switching operations in Southern Railway of British Columbia’s New Westminster Yard as an example of the methodology which can be repeated in other cases as well.

Fully Integrated Quantitative Multiparametric Analysis of Non-Small Cell Lung Cancer at 3-T PET/MRI: Toward One-Stop-Shop Tumor Biological Characterization at the Supervoxel Level.

The aim of this study was to study the feasibility of a fully integrated multiparametric imaging framework to characterize non-small cell lung cancer (NSCLC) at 3-T PET/MRI. An 18F-FDG PET/MRI multiparametric imaging framework was developed and prospectively applied to 11 biopsy-proven NSCLC patients. For each tumor, 12 parametric maps were generated, including PET full kinetic modeling, apparent diffusion coefficient, T1/T2 relaxation times, and DCE full kinetic modeling. Gaussian mixture model-based clustering was applied at the whole data set level to define supervoxels of similar multidimensional PET/MRI behaviors. Taking the multidimensional voxel behaviors as input and the supervoxel class as output, machine learning procedure was finally trained and validated voxelwise to reveal the dominant PET/MRI characteristics of these supervoxels at the whole data set and individual tumor levels. The Gaussian mixture model-based clustering clustering applied at the whole data set level (17,316 voxels) found 3 main multidimensional behaviors underpinned by the 12 PET/MRI quantitative parameters. Four dominant PET/MRI parameters of clinical relevance (PET: k2, k3 and DCE: ve, vp) predicted the overall supervoxel behavior with 97% of accuracy (SD, 0.7; 10-fold cross-validation). At the individual tumor level, these dimensionality-reduced supervoxel maps showed mean discrepancy of 16.7% compared with the original ones. One-stop-shop PET/MRI multiparametric quantitative analysis of NSCLC is clinically feasible. Both PET and MRI parameters are useful to characterize the behavior of tumors at the supervoxel level. In the era of precision medicine, the full capabilities of PET/MRI would give further insight of the characterization of NSCLC behavior, opening new avenues toward image-based personalized medicine in this field.

Echocardiographic phenogrouping by machine learning for risk stratification in the general population.

There is a need for better phenotypic characterization of the asymptomatic stages of cardiac maladaptation. We tested the hypothesis that an unsupervised clustering analysis utilizing echocardiographic indexes reflecting left heart structure and function could identify phenotypically distinct groups of asymptomatic individuals in the general population. We prospectively studied 1407 community-dwelling individuals (mean age, 51.2 years; 51.1% women), in whom we performed clinical and echocardiographic examination at baseline and collected cardiac events on average 8.8 years later. Cardiac phenotypes that were correlated at r > 0.8 were filtered, leaving 21 echocardiographic features, and systolic blood pressure for phenogrouping. We employed hierarchical and Gaussian mixture model-based clustering. Cox regression was used to demonstrate the clinical validity of constructed phenogroups. Unsupervised clustering analyses classified study participants into three distinct phenogroups that differed markedly in echocardiographic indexes. Indeed, cluster 3 had the worst left ventricular (LV) diastolic function (i.e. lowest e' velocity and left atrial (LA) reservoir strain, highest E/e', and LA volume index) and LV remodelling. The phenogroups were also different in cardiovascular risk factor profiles. We observed increase in the risk for incidence of adverse events across phenogroups. In the third phenogroup, the multivariable adjusted risk was significantly higher than the average population risk for major cardiovascular events (51%, P = 0.0028). Unsupervised learning algorithms integrating routinely measured cardiac imaging and haemodynamic data can provide a clinically meaningful classification of cardiac health in asymptomatic individuals. This approach might facilitate early detection of cardiac maladaptation and improve risk stratification.

An artificial bee colony algorithm for mixture model-based clustering

Finite mixture model-based clustering is a popular method to perform unsupervised learning, however the classic approach for parameter estimation, the expectation-maximization algorithm, is quite susceptible to converging to, at times extremely poor, local maxima. Recently, swarm-based algorithms have gained popularity in a number of optimization scenarios. One such swarm-based optimization procedure, the artificial bee colony, incorporates both local searching and multiple solution generating to explore the parameter space. Herein, we propose hybridizing these algorithms for the task of fitting a multivariate Gaussian mixture model. This algorithm’s performance is contrasted with both the traditional approach and another hybrid algorithm on simulated and real data.

A Variational Approximations-DIC Rubric for Parameter Estimation and Mixture Model Selection Within a Family Setting

Mixture model-based clustering has become an increasingly popular data analysis technique since its introduction over fifty years ago, and is now commonly utilized within a family setting. Families of mixture models arise when the component parameters, usually the component covariance (or scale) matrices, are decomposed and a number of constraints are imposed. Within the family setting, model selection involves choosing the member of the family, i.e., the appropriate covariance structure, in addition to the number of mixture components. To date, the Bayesian information criterion (BIC) has proved most effective for model selection, and the expectation-maximization (EM) algorithm is usually used for parameter estimation. In fact, this EM-BIC rubric has virtually monopolized the literature on families of mixture models. Deviating from this rubric, variational Bayes approximations are developed for parameter estimation and the deviance information criteria (DIC) for model selection. The variational Bayes approach provides an alternate framework for parameter estimation by constructing a tight lower bound on the complex marginal likelihood and maximizing this lower bound by minimizing the associated Kullback-Leibler divergence. The framework introduced, which we refer to as VB-DIC, is applied to the most commonly used family of Gaussian mixture models, and real and simulated data are used to compared with the EM-BIC rubric.

Abstract P131: Optimal Phenomapping Strategy to Identify Novel Subgroups of Patients With Type 2 Diabetes and High Cardiovascular Disease Risk

Introduction: Phenomapping can be used to identify subgroups of patients with similar risk factor profiles; however, there is no widely accepted clustering method. We compared 3 phenomapping strategies to identify groups of individuals with diabetes mellitus (DM) and high cardiovascular disease (CVD) risk. Methods: Participants with DM and no baseline CVD from the ACCORD trial (N = 6591) were included in this study. Among 45 mixed data variables, we used an informed variable selection method with optimization of the Wald index to select the top 20 covariates associated with survival. Gaussian mixture models (GMM), Latent Class Analysis (LCA), and finite mixture model-based clustering (FMM) were performed to identify mutually exclusive phenogroups. Bayesian information criterion (BIC) and Dunn index were calculated for each method and compared. The primary composite endpoint was cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. Results: FMM clustering with 3 phenogroups was the optimal clustering strategy as determined by the lowest BIC and highest Dunn index (Figs A-B). Phenogroup 1 had a higher burden of comorbidities and DM complications; phenogroup 2 were older with intermediate CV comorbidity burden, but lowest complications from DM; and phenogroup 3, the largest group, had a lower prevalence of CV-related comorbidities. In adjusted Cox models, over 4.9 years median follow-up, phenogroup 1 and 2 participants (vs. phenogroup 3) had a significantly higher risk of primary composite endpoint (Fig C). In Cox analysis, when group membership was added to 10 optimal predictors of primary composite endpoint, the C -index improved from 0.68 to 0.70 ( P = 0.02). Similar risk patterns were also observed in an external cohort (Look AHEAD trial, N = 4211) with the highest risk of adverse outcome seen in phenogroup 1 (Fig D). Conclusion: Semi-supervised clustering using FMM efficiently identifies phenogroups of patients with DM with distinct clinical characteristics and CVD risk.