Statistical Clustering Algorithm Research Articles

Model of Lexico-Semantic Bonds between Texts for Creating Their Similarity Metrics and Developing Statistical Clustering Algorithm

To solve the problem of text clustering according to semantic groups, we suggest using a model of a unified lexico-semantic bond between texts and a similarity matrix based on it. Using lexico-semantic analysis methods, we can create “term–document” matrices based both on the occurrence frequencies of words and n-grams and the determination of the degrees of nodes in their semantic network, followed by calculating the cosine metrics of text similarity. In the process of the construction of the text similarity matrix using lexical or semantic analysis methods, the cosine of the angle for a vector pair describing such texts will determine the degree of similarity in the lexical or semantic presentation, respectively. Based on the averaging procedure described in this paper, we can obtain a matrix of cosine metric values that describes the lexico-semantic bonds between texts. We propose an algorithm for solving text clustering problems. This algorithm allows one to use the statistical characteristics of the distribution functions of element values in the rows of the cosine metric value matrix in the model of the lexico-semantic bond between documents. In addition, this algorithm allows one to separately describe the matrix of the cosine metric values obtained separately based on the lexical or semantic properties of texts. Our research has shown that the developed model for the lexico-semantic presentation of texts allows one to slightly increase the accuracy of their subsequent clustering. The statistical text clustering algorithm based on this model shows excellent results that are comparable to those of the widely used affinity propagation algorithm. Additionally, our algorithm does not require specification of the degree of similarity for combining vectors into a common cluster and other configuration parameters. The suggested model and algorithm significantly expand the list of known approaches for determining text similarity metrics and their clustering.

Integrated application of prefabricated building construction information based on BIM and RFID technology

Purpose: To solve the problems of low integration accuracy and long integration time of traditional prefabricated construction information integration methods. Method: A method of assembling building information integration based on BIM and RFID technology was proposed. By analyzing the information integration principle of BIM RFID (Building Information Modeling Radio Frequency Identification) technology, starting with rfid technology, we use rfid technology to collect the information of prefabricated building components and obtain the coding information of component data. Experiment: Combining Markov model and fuzzy algorithm, the obtained coding information is preprocessed. According to the processing results, statistical feature clustering algorithm is introduced to integrate the construction information of prefabricated buildings. Result: The precision polyline of the prefabricated building construction information integration method based on BIM and RFID technology showed a steady increase, and it was close to 100% in the later stage. At the same time, the time consumed by this method was within 0.41 s, with high accuracy, high efficiency and high practicability.

Mixed Milk Feeding: A New Approach to Describe Feeding Patterns in the First Year of Life Based on Individual Participant Data from Two Randomised Controlled Trials.

‘Mixed Milk Feeding’ (MMF), whereby infants are fed with both breastmilk and infant formula during the same period, is a common feeding practice. Despite its high prevalence, knowledge regarding MMF practices and their association with (health) outcomes is limited, potentially because MMF behaviours are highly variable and difficult to standardise longitudinally. In this paper, we applied a statistical clustering algorithm on individual infant feeding data collected over the first year of life from two clinical trials: ‘TEMPO’ (n = 855) and ‘Venus’ (n = 539); these studies were conducted in different years and world regions. In TEMPO, more than half of infants were MMF. Four distinct MMF clusters were identified: early exclusive formula feeding (32%), later exclusive formula feeding (25%), long-term MMF (21%), and mostly breastfeeding (22%). The same method applied to ‘Venus’ resulted in comparable clusters, building trust in the robustness of the cluster approach. These results demonstrate that distinct MMF patterns can be identified, which may be applicable to diverse populations. These insights could support the design of future research studying the impact of infant feeding patterns on health outcomes. To standardise this in future research, it is important to establish a unified definition of MMF.

Determination of the background pollution in the Eastern Mediterranean applying a statistical clustering technique

The knowledge of the background air pollution is essential in order to implement abatement strategies for improving air quality. In the present study, a statistical clustering algorithm based on Hidden Markov Models (HMMs) was used to determine the background concentrations of PM10 and O3 at a coastal monitoring site in the Eastern Mediterranean by analyzing an eight-year period of air quality data, from 2011 to 2018. Time series of the deseasonalized daily average concentrations of PM10 and O3 and their diurnal amplitudes were applied to the model to identify the unobserved patterns or hidden states that define the pollution states of the two mentioned air pollutants. The pollution state characterized by the lowest daily average concentration and diurnal amplitudes defined the background pollution level and had the main interest in the current work. The findings revealed that during the studied period, 59% of the total daily PM10 values and 67% of the O3 values corresponded to the background pollution state and the average background concentrations were 14.3 ± 5.2 μg/m3 for PM10 and 48.4 ± 8.2 ppb for O3 respectively. In the case of PM10, anthropogenic activities and Saharan dust events, which define different pollution states, contribute in a decisive way to the PM10 load in the region and play a significant role in the observed annual PM10 levels. Additionally, in the case of O3, anthropogenic activities and regional pollution affect the background levels of the region mainly during summer months. The proposed method is a flexible tool with great potential for estimating the background levels of air pollutants at various monitoring stations.

Application of statistical machine learning clustering algorithms to improve EUR predictions using decline curve analysis in shale-gas reservoirs

A robust machine learning workflow using two well-established statistical clustering algorithms, K-Means Clustering (K-MC) and K-Nearest Neighbors (K-NN) was developed to improve ultimate recovery (EUR) predictions of new wells in shale-gas reservoirs using a decline curve model developed by Duong (2011). These two clustering algorithms can handle big datasets with multiple well parameters with high computational efficiency. Out of a total of 55,623 dry-gas wells from seven shale gas formations, 7631 wells that fit the criteria for Duong's decline curve model were selected in this study for further analysis. K-MC and K-NN were applied to group wells with similar well characteristics using well location, well depth, well length, and production starting year as parameters. Locations of the grouped wells show that the clusters of wells with similar well characteristics identified by both clustering methods scatter over a big area. These findings clearly demonstrate that simply grouping wells by the proximity of physical locations is not a good way of identifying wells with similar well characteristics. The results from this study also suggest that the selection of optimal clustering methods is highly dependent on the shale formation and the method used for EUR prediction. Even with the limited number of well parameters available in the datasets used in this study, the machine learning clustering algorithms managed to improve the EUR prediction accuracy by around 20%. More available well information will allow the machine learning algorithms to capture more of the well characteristics, thereby further increasing the accuracy of EUR predictions.

Is a High Association Between Repetitive Negative Thinking and Negative Affect Predictive of Depressive Symptoms? A Clustering Approach for Experience-Sampling Data

A reciprocal relationship between repetitive negative thinking (RNT) and negative affect (NA) has been found in various types of psychopathology. Recent studies have suggested that the magnitude of this association can vary across time and individuals, which may inform future psychopathology. Here, we explored how these dynamics and interplays are manifested in student and general populations using a statistical clustering algorithm. Across three experience-sampling data sets, our clustering analyses consistently identified two groups of individuals; one group had a higher bidirectional association between RNT and NA (and also higher inertia) than the other group. Furthermore, a prospective analysis revealed that the group with the higher bidirectional association is at risk of developing depressive symptoms during the 3-month follow-up period if they had experienced high levels of NA over the experience-sampling phase. These findings suggest that the dysfunctional affective and cognitive dynamics would be a promising target of preventive intervention.

Asthma clustering methods: a literature-informed application to the children’s health study data

Objective The heterogeneity of asthma has inspired widespread application of statistical clustering algorithms to a variety of datasets for identification of potentially clinically meaningful phenotypes. There has not been a standardized data analysis approach for asthma clustering, which can affect reproducibility and clinical translation of results. Our objective was to identify common and effective data analysis practices in the asthma clustering literature and apply them to data from a Southern California population-based cohort of schoolchildren with asthma. Methods As of January 1, 2020, we reviewed key statistical elements of 77 asthma clustering studies. Guided by the literature, we used 12 input variables and three clustering methods (hierarchical clustering, k-medoids, and latent class analysis) to identify clusters in 598 schoolchildren with asthma from the Southern California Children’s Health Study (CHS). Results Clusters of children identified by latent class analysis were characterized by exhaled nitric oxide, FEV1/FVC, FEV1 percent predicted, asthma control and allergy score; and were predictive of control at two year follow up. Clusters from the other two methods were less clinically remarkable, primarily differentiated by sex and race/ethnicity and less predictive of asthma control over time. Conclusion Upon review of the asthma phenotyping literature, common approaches of data clustering emerged. When applying these elements to the Children’s Health Study data, latent class analysis clusters—represented by exhaled nitric oxide and spirometry measures—had clinical relevance over time.

Statistical hierarchical clustering algorithm for outlier detection in evolving data streams

Anomaly detection is a hard data analysis process that requires constant creation and improvement of data analysis algorithms. Using traditional clustering algorithms to analyse data streams is impossible due to processing power and memory issues. To solve this, the traditional clustering algorithm complexity needed to be reduced, which led to the creation of sequential clustering algorithms. The usual approach is two-phase clustering, which uses online phase to relax data details and complexity, and offline phase to cluster concepts created in the online phase. Detecting anomalies in a data stream is usually solved in the online phase, as it requires unreduced data. Contrarily, producing good macro-clustering is done in the offline phase, which is the reason why two-phase clustering algorithms have difficulty being equally good in anomaly detection and macro-clustering. In this paper, we propose a statistical hierarchical clustering algorithm equally suitable for both detecting anomalies and macro-clustering. The proposed algorithm is single-phased and uses statistical inference on the input data stream, resulting in statistical distributions that are constantly updated. This makes the classification adaptable, allowing agglomeration of outliers into clusters, tracking population evolution, and to be used without knowing the expected number of clusters and outliers. The proposed algorithm was tested against typical clustering algorithms, including two-phase algorithms suitable for data stream analysis. A number of typical test cases were selected, to show the universality and qualities of the proposed clustering algorithm.

Virtual Patients: An Enabling Technology for Multivariable Control of Biomedical Systems

This paper presents the development of virtual patients to enable the simulation evaluation and assessment of multivariable control algorithms for biomedical systems. The virtual patients are generated by fitting the parameters of the models to clinical experimental data, followed by the estimation of the multivariate distribution of the actual patient parameters. The estimated multivariate distribution is then incorporated with constraints to ensure the sampling of synthetic virtual patients conforms to the actual patient parameter bounds. The sampled synthetic virtual patients are analyzed through multivariate statistical techniques and data clustering algorithms to prune out virtual subjects with similar characteristics or unrealistic dynamics, yielding a virtual patient population that is diverse and with individually distinct characteristics. The generated virtual patient population is used to evaluate multivariable nonlinear and adaptive control algorithms for insulin dosing in people with Type 1 diabetes.

Economic Evaluation Model of Freight Distribution Management in Maritime Port

Mao, Y. and Gan, S., 2019. Economic evaluation model of freight distribution management in maritime port. In: Guido-Aldana, P.A. and Mulahasan, S. (eds.), Advances in Water Resources and Exploration. Journal of Coastal Research, Special Issue No. 93, pp. 1059–1065. Coconut Creek (Florida), ISSN 0749-0208.The economic benefit evaluation model of seaborne port cargo distributed transportation management is proposed based on big data's integrated dispatching of resources, and the explanatory variable model and control variable model of the economic benefit evaluation of seaborne port cargo distributed transportation management is analyzed by using quantitative analysis method. The distribution characteristics of economic statistical information and the characteristics of association rules in the management of seaborne port cargo distributed transportation are taken into account. Rough evaluation set and economic statistical information distribution structure model of cargo distributed transportation management are constraint parameters, so fuzzy subspace scheduling model is constructed to design the control model for economic benefit evaluation of cargo distributed transportation management in seaborne ports. Based on the quantitative regression analysis of the intervention factors of the economic statistical participants in the management of goods distributed transportation, the benefit evaluation theory and regression analysis method of the distributed transportation of goods in the seaborne port are used to carry out the management of the distributed transportation of the goods in the seaborne port, and the economic statistical information fusion and clustering algorithm of the distributed transportation management of the goods in the seaborne port are combined to realize the economic benefit evaluation of the distributed transportation management of the goods in the seaborne port. The simulation results show that the economic benefit evaluation of seaborne port cargo distributed transportation management is accurate, and the adaptive evaluation ability and confidence of the economic evaluation quality benefit coefficient of seaborne port cargo distributed transportation management are strong.

In Situ Additive Manufacturing Process Monitoring With an Acoustic Technique: Clustering Performance Evaluation Using K-Means Algorithm

Additive manufacturing (AM) is based on layer-by-layer addition of materials. It gives design flexibility and potential to decrease costs and manufacturing lead time. Because the AM process involves incremental deposition of materials, it provides unique opportunities to investigate the material quality as it is deposited. Development of in situ monitoring methodologies is a vital part of the assessment of process performance and understanding of defects formation. In situ process monitoring provides the capability for early detection of process faults and defects. Due to the sensitivity of AM processes to different factors such as laser and material properties, any changes in aspects of the process can potentially have an impact on the part quality. As a result, in-process monitoring of AM is crucial to assure the quality, integrity, and safety of AM parts. There are various sensors and techniques that have been used for in situ process monitoring. In this work, acoustic signatures were used for in situ monitoring of the metal direct energy deposition (DED) AM process operating under different process conditions. Correlations were demonstrated between metrics and various process conditions. Demonstrated correlation between the acoustic signatures and the manufacturing process conditions shows the capability of acoustic technique for in situ monitoring of the additive manufacturing process. To identify the different process conditions, a new approach of K-means statistical clustering algorithm is used for the classification of different process conditions, and quantitative evaluation of the classification performance in terms of cohesion and isolation of the clusters. The identified acoustic signatures, quantitative clustering approach, and the achieved classification efficiency demonstrate potential for use in in situ acoustic monitoring and quality control for the additive manufacturing process.

Faces of Joblessness in Australia: An Anatomy of Employment Barriers Using Household Data

Australia's economy and labour market have escaped a dramatic downturn following the global financial economic crisis. Yet, a substantial share of working-age Australians either were not working or worked only to a limited extent as the global recovery gathered pace between 2013 and 2014. 18% were without employment during an entire year; a further 6% had weak labour-market attachment, e.g. working only a fraction of the year. This paper extends a method proposed by Fernandez et al. (2016) to measure and visualise employment barriers of individuals with weak labour-market attachment using household micro-data. It first develops indicators to quantify employment obstacles under three headings: (i) work-related capabilities, (ii) incentives, and (iii) employment opportunities. A novelty in this paper is a statistical procedure for calibrating the definition of barriers in a way that maximises their explanatory power in predicting employment outcomes. A statistical clustering algorithm then identify groups with similar combinations of barriers. The resulting typology provides insights on the most pressing policy priorities in supporting different groups into employment in Australia. We identify seven distinct groups, each calling for a specific flavour of activation and employment-support policies. The most common employment obstacles are limited work experience, low skills and poor health. Financial disincentives, care responsibilities and scarce job opportunities are less widespread overall but were important barriers for some groups. Almost one third of jobless or low-intensity workers face three or more simultaneous barriers, highlighting the limits of policy approaches that focus on subsets of these employment obstacles in isolation.

On the effect of model parameters on forecast objects

Abstract. Many physics-based numerical models produce a gridded, spatial field of forecasts, e.g., a temperature map. The field for some quantities generally consists of spatially coherent and disconnected objects. Such objects arise in many problems, including precipitation forecasts in atmospheric models, eddy currents in ocean models, and models of forest fires. Certain features of these objects (e.g., location, size, intensity, and shape) are generally of interest. Here, a methodology is developed for assessing the impact of model parameters on the features of forecast objects. The main ingredients of the methodology include the use of (1) Latin hypercube sampling for varying the values of the model parameters, (2) statistical clustering algorithms for identifying objects, (3) multivariate multiple regression for assessing the impact of multiple model parameters on the distribution (across the forecast domain) of object features, and (4) methods for reducing the number of hypothesis tests and controlling the resulting errors. The final output of the methodology is a series of box plots and confidence intervals that visually display the sensitivities. The methodology is demonstrated on precipitation forecasts from a mesoscale numerical weather prediction model.

Statistical clustering of parametric maps from dynamic contrast enhanced MRI and an associated decision tree model for non-invasive tumour grading of T1b solid clear cell renal cell carcinoma.

To apply a statistical clustering algorithm to combine information from dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) into a single tumour map to distinguish high-grade from low-grade T1b clear cell renal cell carcinoma (ccRCC). This prospective, Institutional Review Board -approved, Health Insurance Portability and Accountability Act -compliant study included 18 patients with solid T1b ccRCC who underwent pre-surgical DCE MRI. After statistical clustering of the parametric maps of the transfer constant between the intravascular and extravascular space (K trans ), rate constant (K ep ) and initial area under the concentration curve (iAUC) with a fuzzy c-means (FCM) algorithm, each tumour was segmented into three regions (low/medium/high active areas). Percentages of each region and tumour size were compared to tumour grade at histopathology. A decision-tree model was constructed to select the best parameter(s) to predict high-grade ccRCC. Seven high-grade and 11 low-grade T1b ccRCCs were included. High-grade histology was associated with higher percent high active areas (p = 0.0154) and this was the only feature selected by the decision tree model, which had a diagnostic performance of 78% accuracy, 86% sensitivity, 73% specificity, 67% positive predictive value and 89% negative predictive value. The FCM integrates multiple DCE-derived parameter maps and identifies tumour regions with unique pharmacokinetic characteristics. Using this approach, a decision tree model using criteria beyond size to predict tumour grade in T1b ccRCCs is proposed. • Tumour size did not correlate with tumour grade in T1b ccRCC. • Tumour heterogeneity can be analysed using statistical clustering via DCE-MRI parameters. • High-grade ccRCC has a larger percentage of high active area than low-grade ccRCCs. • A decision-tree model offers a simple way to differentiate high/low-grade ccRCCs.

Background based Gaussian mixture model lesion segmentation in PET.

Quantitative (18)F-fluorodeoxyglucose positron emission tomography is limited by the uncertainty in lesion delineation due to poor SNR, low resolution, and partial volume effects, subsequently impacting oncological assessment, treatment planning, and follow-up. The present work develops and validates a segmentation algorithm based on statistical clustering. The introduction of constraints based on background features and contiguity priors is expected to improve robustness vs clinical image characteristics such as lesion dimension, noise, and contrast level. An eight-class Gaussian mixture model (GMM) clustering algorithm was modified by constraining the mean and variance parameters of four background classes according to the previous analysis of a lesion-free background volume of interest (background modeling). Hence, expectation maximization operated only on the four classes dedicated to lesion detection. To favor the segmentation of connected objects, a further variant was introduced by inserting priors relevant to the classification of neighbors. The algorithm was applied to simulated datasets and acquired phantom data. Feasibility and robustness toward initialization were assessed on a clinical dataset manually contoured by two expert clinicians. Comparisons were performed with respect to a standard eight-class GMM algorithm and to four different state-of-the-art methods in terms of volume error (VE), Dice index, classification error (CE), and Hausdorff distance (HD). The proposed GMM segmentation with background modeling outperformed standard GMM and all the other tested methods. Medians of accuracy indexes were VE <3%, Dice >0.88, CE <0.25, and HD <1.2 in simulations; VE <23%, Dice >0.74, CE <0.43, and HD <1.77 in phantom data. Robustness toward image statistic changes (±15%) was shown by the low index changes: <26% for VE, <17% for Dice, and <15% for CE. Finally, robustness toward the user-dependent volume initialization was demonstrated. The inclusion of the spatial prior improved segmentation accuracy only for lesions surrounded by heterogeneous background: in the relevant simulation subset, the median VE significantly decreased from 13% to 7%. Results on clinical data were found in accordance with simulations, with absolute VE <7%, Dice >0.85, CE <0.30, and HD <0.81. The sole introduction of constraints based on background modeling outperformed standard GMM and the other tested algorithms. Insertion of a spatial prior improved the accuracy for realistic cases of objects in heterogeneous backgrounds. Moreover, robustness against initialization supports the applicability in a clinical setting. In conclusion, application-driven constraints can generally improve the capabilities of GMM and statistical clustering algorithms.

Lithofacies classification in the Marcellus Shale by applying a statistical clustering algorithm to petrophysical and elastic well logs

Interpreting shale lithofacies is an important step in identifying productive zones in the Marcellus Shale gas play; the target reservoir can be less than 3 m (10 ft) thick with overlying shales that appear similar on certain petrophysical well logs and in the core. However, these nonreservoir-quality shales contain widely varying organic and mechanical properties. To distinguish between reservoir- and nonreservoir-quality shale facies, this classification method applies a pattern-recognition algorithm, expectation maximization, to a set of commonly available petrophysical and elastic well logs to create 1D shale facies models along the wellbore. The statistical algorithm classifies the petrophysical and elastic well-log data into defined facies using the theory that each well log contains a distribution of Gaussian curves for each facies. The method is applied to 12 wells across Pennsylvania and northern West Virginia, and it is able to discriminate between shale facies based on organic content and brittleness, characteristics which are not always evident in core. To verify the geologic accuracy of the facies models, the results are compared with the core data, well logs, mudlogs, and regional stratigraphic studies when available. In addition, we were able to designate a lithofacies-defined top to the Marcellus Formation in parts of the Appalachian Basin where the gradational contact with the overlying Mahantango Formation is poorly defined on petrophysical logs.

An Algorithm for Imputing Missing Values in Microarray Gene Expression Data

In recent years, rapid developments in genomics and proteomics have generated a large amount of biological data. Dealing with such huge data has become extremely challenging with traditional data analysis techniques. Bioinformatics is the interdisciplinary science of interpreting biological data that uses the tools and techniques of information technology and computer science. However, gene expressions generated by the high-throughput microarray experiments often contain missing values, which significantly affect the performance of subsequent statistical analysis and clustering algorithms. So there is a great need for estimating or imputing these missing values as accurately as possible. In general the missing values could be imputed by means of various methods namely ignoring the tuple, using the attribute mean to fill the missing value, using a global constant to fill in the missing value. In this paper a new approach called JAD (Java Application Development) imputation is proposed for missing values that can be estimated more accurately. The results show that the proposed method called JAD imputation provides a better solution to complete the microarray gene expression data.

Synchronous low frequency earthquakes and implications for deep San Andreas Fault slip

Low Frequency Earthquakes (LFEs) are slip events that occur repeatedly at source locations within the lower crust. LFEs, and the associated seismic broadcast known as tremor, have been observed in a diverse array of tectonic environments. Here we develop a suite of statistical tools to conduct a systematic study of the spatial and temporal correlations of the event occurrence patterns of the 88 LFE sources beneath the greater Parkfield section of the San Andreas Fault. We first examine correlations in the occurrence patterns on long time scales to show that the regions to the north and south of Parkfield behave independently. We next use the cumulative event signatures of each source to characterize the individual occurrence patterns on shorter time scales. Through application of a statistical clustering algorithm, we demonstrate that individual LFE sources form spatially coherent clusters that may represent localized elastic structures or asperities on the deep fault interface. We conclude by examining the fine-scale features of the event rates within the LFE occurrence patterns. Through quantitative comparison to analogous laboratory shear experiments on granular, fault gouge-like materials, we infer that the distinctive features of LFE occurrence patterns reflect variations in the in-situ stress and frictional conditions at the individual LFE source locations. These observations provide a framework to understand the spatial and temporal diversity of fault slip that occurs within the lower crust beneath Parkfield and that may influence seismic hazard in the region.

Dynamic Reserve Zones for Day-Ahead Unit Commitment With Renewable Resources

As more non-dispatchable renewable resources are integrated into the grid, it will become increasingly difficult to predict the transfer capabilities and the network congestion. At the same time, renewable resources require operators to acquire more operating reserves. With today's deterministic reserve requirements unable to ensure optimal reserve locations, improvements to reserve policies are needed to ensure reserve deliverability while maintaining a reliable system at least cost. This paper proposes a daily reserve zone determination procedure, which is able to reflect system operating conditions by utilizing probabilistic power flows. A statistical clustering algorithm is used to cluster buses together to produce the zones; the algorithm uses a centrality measurement, which is based on weighted power transfer distribution factors. The proposed method is validated by testing it on a modified IEEE 118-bus system for multiple days; the proposed method is compared against existing reserve zone partitioning procedures. While the proposed reserve zone determination method is a heuristic, it is shown to be effective and it is a computationally tractable method. The proposed method can be used on its own and can be used along with stochastic programming techniques that implicitly determine reserves.

BANKING BUSINESS MODELS IN UKRAINIAN BANKING SYSTEM

The purpose of the paper is to work out and characterize bank business models that are formed in Ukraine. Methodology. Our research we will spend among banks that are functioning on the Ukrainian financial market and are not on the stage of liquidation, so the sample under study in our work is comprised of 131 banks which are different in their ownership structure and size. The core of the methodology is a statistical clustering algorithm that allows identifying the groups of banks (clusters) with similar business models as banks with similar business model strategies have made similar choices regarding the composition of their assets and liabilities. The cluster analyses were taken on the base of seven chosen indicators: bank loans, bank liabilities, enterprise loans, enterprise liabilities, household loans, household liabilities and trading assets. Results. The traditional business model of bank is worked out. The bank business models that are functioning in Ukraine are identified on the base of cluster analyses using balance sheet characteristics of 131 Ukrainian banks. We find that in Ukraine were formed three types of bank business models: “Focused retail”, “Diversified retail” and “Corporative retail”. The description of each model is given. Practical implications. More detailed research of distinguished models allows not only to find out the main advantages and disadvantages of each bank model, but also the main problems that follow the development of Ukrainian banking sector. Identifying of bank models and their studying simplifies searching and elaboration of regulatory instruments as there is a two-way causation between regulation and bank business models. This implies a symbiotic relationship between regulation and bank business models: business models respond to regulation which in turn responds to the evolution of new business models. Value/originality. Such survey is conducted at the first time among Ukrainian banks. The main differences which are between European and Ukrainian bank business models are marked out. Recommendations concerning improving of bank business models in Ukraine are given.