Unsupervised Machine Learning Methods Research Articles

Open science perspectives on machine learning for the identification of careless responding: A new hope or phantom menace?

AbstractPowerful methods for identifying careless respondents in survey data are not just important to ensure the validity of subsequent data analyses, they are also instrumental for studying the psychological processes that drive humans to respond carelessly. Conversely, a deeper understanding of the phenomenon of careless responding enables the development of improved methods for the identification of careless respondents. While machine learning has gained substantial attention and popularity in many scientific fields, it is largely unexplored for the detection of careless responding. On the one hand, machine learning algorithms can be highly powerful tools due to their flexibility. On the other hand, science based on machine learning has been criticized in the literature for a lack of reproducibility. We assess the potential and the pitfalls of machine learning approaches for identifying careless respondents from an open science perspective. In particular, we discuss possible sources of reproducibility issues when applying machine learning in the context of careless responding, and we give practical guidelines on how to avoid them. Furthermore, we illustrate the high potential of an unsupervised machine learning method for the identification of careless respondents in a proof‐of‐concept simulation experiment. Finally, we stress the necessity of building an open data repository with labeled benchmark data sets, which would enable the evaluation of methods in a more realistic setting and make it possible to train supervised learning methods. Without such a data repository, the true potential of machine learning for the identification of careless responding may fail to be unlocked.

Disentangling boredom from depression using the phenomenology and content of involuntary autobiographical memories

Recurrent involuntary autobiographical memories (IAMs) are memories retrieved unintentionally and repetitively. We examined whether the phenomenology and content of recurrent IAMs could differentiate boredom and depression, both of which are characterized by affective dysregulation and spontaneous thought. Participants (n = 2484) described their most frequent IAM and rated its phenomenological properties (e.g., valence). Structural topic modeling, a method of unsupervised machine learning, identified coherent content within the described memories. Boredom proneness was positively correlated with depressive symptoms, and both boredom proneness and depressive symptoms were correlated with more negative recurrent IAMs. Boredom proneness predicted less vivid recurrent IAMs, whereas depressive symptoms predicted more vivid, negative, and emotionally intense ones. Memory content also diverged: topics such as relationship conflicts were positively predicted by depressive symptoms, but negatively predicted by boredom proneness. Phenomenology and content in recurrent IAMs can effectively disambiguate boredom proneness from depressive symptoms in a large sample of undergraduate students from a racially diverse university.

Improving Reservoir Characterization and Prediction Via Machine Learning-Driven Integration of Subseismic Geologic Concepts, Geophysical Attributes, and Wells

Summary Reservoir dynamic performance is affected by subseismic (below the resolution of seismic imaging) geologic features such as submeter thin barriers and baffles, or high-permeability streaks. Here, we present an application of machine learning that has the potential to improve and significantly accelerate the workflow for generating reservoir realizations and scenarios with relevant subseismic features consistent with both well-log and 3D/4D seismic data and/or attributes. We leverage unsupervised machine learning methods, in particular, a modification of diffusion probabilistic modeling (DPM).

Acceleration of Molecular Simulations by Parametric Time-Lagged tSNE Metadynamics.

The potential of molecular simulations is limited by their computational costs. There is often a need to accelerate simulations using some of the enhanced sampling methods. Metadynamics applies a history-dependent bias potential that disfavors previously visited states. To apply metadynamics, it is necessary to select a few properties of the system─collective variables (CVs) that can be used to define the bias potential. Over the past few years, there have been emerging opportunities for machine learning and, in particular, artificial neural networks within this domain. In this broad context, a specific unsupervised machine learning method was utilized, namely, parametric time-lagged t-distributed stochastic neighbor embedding (ptltSNE) to design CVs. The approach was tested on a Trp-cage trajectory (tryptophan cage) from the literature. The trajectory was used to generate a map of conformations, distinguish fast conformational changes from slow ones, and design CVs. Then, metadynamic simulations were performed. To accelerate the formation of the α-helix, we added the α-RMSD collective variable. This simulation led to one folding event in a 350 ns metadynamics simulation. To accelerate degrees of freedom not addressed by CVs, we performed parallel tempering metadynamics. This simulation led to 10 folding events in a 200 ns simulation with 32 replicas.

Leveraging machine learning for taxonomic classification of emerging astroviruses.

Astroviruses are a family of genetically diverse viruses associated with disease in humans and birds with significant health effects and economic burdens. Astrovirus taxonomic classification includes two genera, Avastrovirus and Mamastrovirus. However, with next-generation sequencing, broader interspecies transmission has been observed necessitating a reexamination of the current host-based taxonomic classification approach. In this study, a novel taxonomic classification method is presented for emergent and as yet unclassified astroviruses, based on whole genome sequence k-mer composition in addition to host information. An optional component responsible for identifying recombinant sequences was added to the method's pipeline, to counteract the impact of genetic recombination on viral classification. The proposed three-pronged classification method consists of a supervised machine learning method, an unsupervised machine learning method, and the consideration of host species. Using this three-pronged approach, we propose genus labels for 191 as yet unclassified astrovirus genomes. Genus labels are also suggested for an additional eight as yet unclassified astrovirus genomes for which incompatibility was observed with the host species, suggesting cross-species infection. Lastly, our machine learning-based approach augmented by a principal component analysis (PCA) analysis provides evidence supporting the hypothesis of the existence of human astrovirus (HAstV) subgenus of the genus Mamastrovirus, and a goose astrovirus (GoAstV) subgenus of the genus Avastrovirus. Overall, this multipronged machine learning approach provides a fast, reliable, and scalable prediction method of taxonomic labels, able to keep pace with emerging viruses and the exponential increase in the output of modern genome sequencing technologies.

Experimental investigation on acoustic emission precursor of rockburst based on unsupervised machine learning method

The key to achieving rockburst warning lies in the understanding of rockburst precursors. Considering the correlation characteristics of rockburst acoustic emission (AE) parameters, a self-organizing map neural network (SOMNN) based method for rockburst precursor inversion was proposed. The feature of this method lies in a cyclic data segmentation iteration process based on the thinking of “interference signal screening”, “key signal extraction”, and “precursor signal inversion”. The rationality of this method has been verified in three groups of rockburst experiments. The results revealed that rockburst AE precursor signals consist of a series of signals characterized by long duration, high energy, low average frequency, high energy amplitude, and low peak frequency. Subsequently, potential value in long term rockburst warning of the precursor obtained in this study was shown via the comparison of conventional precursors. Finally, a preliminary interpretation for rockburst precursor was proposed under the framework of AE parameters physical significance, and it is revealed that AE precursor signals are likely linked to the creation of large-scale tensile cracks before rockburst.

Multi-well clustering and inverse modeling-based approaches for exploring geometry, petrophysical, and hydrogeological parameters of the Quaternary aquifer system around Debrecen area, Hungary.

This research aims to explore the application of an unsupervised machine learning and inverse modeling-based methods to map the aquifers geometry and investigate the petrophysical and hydrogeological parameters of the Quaternary aquifer system around the Debrecen area, Eastern Hungary. The study utilized a limited well-logs, including spontaneous potential, natural gamma ray, and normal resistivity logs. The k-means clustering technique is applied to identify the distribution of lithological facies within the formerly identified basin-scale hydrostratigraphical units of coarsening upward, alluvial, valley incision, and Late Miocene deposits. Based on the mathematical and geological considerations, the analysis revealed three main clusters (C1, C2, and C3), representing different lithofacies of shale, shaly sand, and sand and gravel. The result of the cluster analysis is further validated with a surface geophysical survey using the vertical electrical sounding (VES) technique. Furthermore, an inverse-modeling-based approach using Csókás method is employed to detect the vertical and horizontal distribution of the hydraulic conductivity along these units. Csókás model is empirically modified from the Kozeny-Carman equation that suites the unconsolidated freshwater-bearing units. The results indicated that the petrophysical and hydrogeological parameters widely ranged with the hydraulic conductivity between almost zero in the shaly layers to more than 21.5 m/d in the sandy and gravely layers. However, the valley incision unit aquifers showed a more uniform distribution of hydraulic conductivity. Based on these results, the Quaternary aquifers in the Debrecen area are classified as moderate to highly productive and ideal for groundwater development. The applied methodology contributed to understanding the complexity of the hydrogeological conditions, providing a robust approach to characterize the heterogeneous groundwater systems.

Epigenetic Control of Adamantinomatous Craniopharyngiomas.

Studies addressing the methylation pattern in adamantinomatous craniopharyngioma (ACP) are lacking. To identify methylation signatures in ACPs regarding clinical presentation and outcome. Clinical and pathology data were collected from 35 patients with ACP (54% male; 18.1 years [2-68]). CTNNB1 mutations and methylation profile (MethylationEPIC/Array-Illumina) were analyzed in tumoral DNA. Unsupervised machine learning analysis of this comprehensive methylome sample was achieved using hierarchical clustering and multidimensional scaling. Statistical associations between clusters and clinical features were achieved using the Fisher test and global biological process interpretations were aided by Gene Ontology enrichment analyses. Two clusters were revealed consistently by all unsupervised methods (ACP-1: n = 18; ACP-2: n = 17) with strong bootstrap statistical support. ACP-2 was enriched by CTNNB1 mutations (100% vs 56%, P = .0006), hypomethylated in CpG island, non-CpG Island sites, and globally (P < .001), and associated with greater tumor size (24.1 vs 9.5 cm3, P = .04). Enrichment analysis highlighted pathways on signaling transduction, transmembrane receptor, development of anatomical structures, cell adhesion, cytoskeleton organization, and cytokine binding, and cell type-specific biological processes as regulation of oligodendrocytes, keratinocyte, and epithelial cells differentiation. Two clusters of patients with ACP were consistently revealed by unsupervised machine learning methods, with one of them significantly hypomethylated, enriched by CTNNB1 mutated ACPs, and associated with increased tumor size. Enrichment analysis reinforced pathways involved in tumor proliferation and in cell-specific tumoral microenvironment.

Tremor clustering reveals pre-eruptive signals and evolution of the 2021 Geldingadalir eruption of the Fagradalsfjall Fires, Iceland

Analyzing seismic data in a timely manner is essential for potential eruption forecasting and early warning in volcanology. Here, we demonstrate that unsupervised machine learning methods can automatically uncover hidden details from the continuous seismic signals recorded during Iceland’s 2021 Geldingadalir eruption. By pinpointing the eruption’s primary phases, including periods of unrest, ongoing lava extrusion, and varying lava fountaining intensities, we can effectively chart its temporal progress. We detect a volcanic tremor sequence three days before the eruption, which may signify impending eruptive activities. Moreover, the discerned seismicity patterns and their temporal changes offer insights into the shift from vigorous outflows to lava fountaining. Based on the extracted patterns of seismicity and their temporal variations we propose an explanation for this transition. We hypothesize that the emergence of episodic tremors in the seismic data in early May could be related to an increase in the discharge rate in late April.

Clustering honey samples with unsupervised machine learning methods using FTIR data.

This study utilizes Fourier transform infrared (FTIR) data from honey samples to cluster and categorize them based on their spectral characteristics. The aim is to group similar samples together, revealing patterns and aiding in classification. The process begins by determining the number of clusters using the elbow method, resulting in five distinct clusters. Principal Component Analysis (PCA) is then applied to reduce the dataset's dimensionality by capturing its significant variances. Hierarchical Cluster Analysis (HCA) further refines the sample clusters. 20% of the data, representing identified clusters, is randomly selected for testing, while the remainder serves as training data for a deep learning algorithm employing a multilayer perceptron (MLP). Following training, the test data are evaluated, revealing an impressive 96.15% accuracy. Accuracy measures the machine learning model's ability to predict class labels for new data accurately. This approach offers reliable honey sample clustering without necessitating extensive preprocessing. Moreover, its swiftness and cost-effectiveness enhance its practicality. Ultimately, by leveraging FTIR spectral data, this method successfully identifies similarities among honey samples, enabling efficient categorization and demonstrating promise in the field of spectral analysis in food science.

Unsupervised machine learning of virus dispersion indoors

This paper concerns analyses of virus droplet dynamics resulting from coughing events within a confined environment using, as an example, a typical cruiser's cabin. It is of paramount importance to be able to comprehend and predict droplet dispersion patterns within enclosed spaces under varying conditions. Numerical simulations are expensive and difficult to perform in real-time situations. Unsupervised machine learning methods are proposed to study droplet dispersion patterns. Data from multi-phase computational fluid dynamics simulations of coughing events at different flow rates are utilized with an unsupervised learning algorithm to identify prevailing trends based on the distance traveled by the droplets and their sizes. The algorithm determines optimal clustering by introducing novel metrics such as the Clustering Dominance Index and Uncertainty. Our analysis revealed the existence of three distinct stages for droplet dispersion during a coughing event, irrespective of the underlying flow rates. An initial stage where all droplets disperse homogeneously, an intermediate stage where larger droplets overtake the smaller ones, and a final stage where the smaller droplets overtake the larger ones. This is the first time computational fluid dynamics is coupled with unsupervised learning to study particles' dispersion and understand their dynamic behavior.

Radiomics-based machine learning model to phenotype hip involvement in ankylosing spondylitis: a pilot study.

Hip involvement is an important reason of disability in patients with ankylosing spondylitis (AS). Unveiling the potential phenotype of hip involvement in AS remains an unmet need to understand its biological mechanisms and improve clinical decision-making. Radiomics, a promising quantitative image analysis method that had been successfully used to describe the phenotype of a wide variety of diseases, while it was less reported in AS. The objective of this study was to investigate the feasibility of radiomics-based approach to profile hip involvement in AS. A total of 167 patients with AS was included. Radiomic features were extracted from pelvis MRI after image preprocessing and feature engineering. Then, we performed unsupervised machine learning method to derive radiomics-based phenotypes. The validation and interpretation of derived phenotypes were conducted from the perspectives of clinical backgrounds and MRI characteristics. The association between derived phenotypes and radiographic outcomes was evaluated by multivariable analysis. 1321 robust radiomic features were extracted and four biologically distinct phenotypes were derived. According to patient clinical backgrounds, phenotype I (38, 22.8%) and II (34, 20.4%) were labelled as high-risk while phenotype III (24, 14.4%) and IV (71, 42.5%) were at low risk for hip involvement. Consistently, the high-risk phenotypes were associated with higher prevalence of MRI-detected lesion than the low-risk. Moreover, phenotype I had significant acute inflammation signs than phenotype II, while phenotype IV was enthesitis-predominant. Importantly, the derived phenotypes were highly predictive of radiographic outcomes of patients, as the high-risk phenotypes were 3 times more likely to have radiological hip lesion than the low-risk [27 (58.7%) vs 16 (28.6%); adjusted odds ratio (OR) 2.95 (95% CI 1.10, 7.92)]. We confirmed for the first time, the clinical actionability of profiling hip involvement in AS by radiomics method. Four distinct phenotypes of hip involvement in AS were identified and importantly, the high-risk phenotypes could predict structural damage of hip involvement in AS.

DEVELOPMENT OF WIRELESS SENSOR NETWORKS EFFICIENT CLUSTERIZATION METHOD

Wireless sensor networks are a critical element of the Internet of Things and are widely used in various fields. Effective clustering is key to energy efficiency and network stability in wireless sensor networks. Optimize the density of clusters, their number and size to increase the stability and service life of the network – It is one of the tasks of clustering wireless sensor networks. Another problem – effectively organize the network topology in order to balance the load and extend the network lifetime. Clustering has proven to be an effective approach for organizing a network into a cohesive hierarchy. However, there are several key issues that affect the practical application of clustering methods in sensor network applications. These issues include the rationale for developing different clustering approaches, classification of proposed approaches based on their goals and design principles, and problems associated with clustering. K-means algorithm is one of the unsupervised machine learning methods for clustering tasks and is widely used in wireless sensor networks. This article proposes a modified K-means clustering algorithm that eliminates the shortcomings of the basic K-means algorithm. The modified algorithm increases network lifetime, reduces power consumption and improves network stability.

Mathematical enhancement of unsupervised machine learning algorithms for optimal lifeline donor knowledge extraction

Lifeline Donor groups are crucial in the field of humanitarian aid because they provide lifesaving aid in times of crisis. For effective resource management and aid coordination, it is crucial to draw relevant insights from donor data in a timely manner. Scalability, precision, and flexibility in responding to changing donor dynamics are all areas where traditional techniques of knowledge extraction fall short. This study offers a fresh method for optimizing knowledge extraction from Lifeline Donor databases by utilizing improved unsupervised machine learning methods. In order to extract, categorize, and prioritize donor knowledge from disparate data sources, this research presents a multi-faceted framework that includes cutting-edge machine learning algorithms. By merging cutting-edge developments in NLP and data mining, the proposed framework improves upon classic clustering and topic modelling methods. The program is able to capture subtle semantic links and underlying patterns in donor communication data by using methods like word embedding models and graph-based clustering.

Unsupervised machine learning for disease prediction: a comparative performance analysis using multiple datasets

PurposeDisease risk prediction poses a significant and growing challenge in the medical field. While researchers have increasingly utilised machine learning (ML) algorithms to tackle this issue, supervised ML methods remain dominant. However, there is a rising interest in unsupervised techniques, especially in situations where data labels might be missing — as seen with undiagnosed or rare diseases. This study delves into comparing unsupervised ML models for disease prediction.MethodsThis study evaluated the efficacy of seven unsupervised algorithms on 15 datasets, including those of heart failure, diabetes, and breast cancer. It used six performance metrics for this comparison. They are Adjusted Rand Index, Adjusted Mutual Information, Homogeneity, Completeness, V-measure and Silhouette Coefficient.ResultsAmong the seven unsupervised ML methods, the DBSCAN (Density-based spatial clustering of applications with noise) showed the best performance most times (31), followed by the Bayesian Gaussian Mixture (18) and Divisive clustering (15). No single model consistently outshined others across every dataset and metric. The study emphasises the crucial role of model and performance measure selections based on application-specific needs. For example, DBSCAN excels in Homogeneity, Completeness and V-measure metrics. Conversely, the Bayesian Gaussian Mixture is good in the Adjusted R and Index metric. The codes used in this study can be found at https://github.com/haohuilu/unsupervisedml/.ConclusionThis research contributes deeper insights into the unsupervised ML applications in healthcare and encourages further investigations into model selection. Subsequent studies could harness genuine disease records for a more nuanced comparison and evaluation of models.

The added diagnostic value of 18 F-FDG PET/CT radiomic analysis in multiple myeloma patients with negative visual analysis.

A small number of patients diagnosed with multiple myeloma (MM) by bone marrow aspiration reported as being disease-free on 18 F-FDG PET/CT. We aim to evaluate the diagnostic value of radiomics approach in patients with MM who were negative by visual analysis. Thirty-three patients judged negative by visual analysis were assigned to the MM group. Contemporaneous 31 disease-free patients served as the control group. 70% of the whole data set was used as training set (23 from MM group and 22 from control group) and 30% as testing set (10 from MM group and 9 from control group). Axial skeleton volumes were automatically segmented and high-dimensional imaging features were extracted from PET and CT. The unsupervised machine learning method was used to filter and reduce the dimensions of the extracted features. Random forest was used to construct the prediction model and then validated with 10-fold cross-validation and evaluated on the independent testing set. One thousand seven hundred two quantitative features were extracted from PET and CT. Of those, three first-order and one high-order imaging features were uncorrelated. With the cross-validation on the training group, the sensitivity, specificity, accuracy and area under the curve of random forest were 0.850, 0.792, 0.818 and 0.894, respectively. On the independent testing set, the accuracy of the model was 0.850 and the area under the curve was 0.909. Radiomic analysis based on 18 F-FDG PET/CT using machine learning model provides a quantitative, objective and efficient mechanism for diagnosing patients with MM who were negative by visual analysis.

Unsupervised data-driven method for damage localization using guided waves

Researchers have recently gained interest in ultrasonic guided waves (UGWs) for structural health monitoring (SHM). This interest has mainly grown thanks to the high sensitivity of UGWs to changes in the mechanical and material properties of the medium they travel in. Among the several types of UGWs, Lamb waves (LWs) have been successfully employed to detect, localize and quantify damage in thin-walled structures. To date, localization has been mainly performed through tomographic algorithms. Although those algorithms represent consolidated methods, they come with unsolved issues, such as artifacts generation in damage probability maps and strong dependency on the network of sensors installed to excite and sense diagnostic signals. Hence, more recently, supervised data-driven approaches have been propounded in the literature for LW-based damage diagnosis. Despite the good performance of such methods, a strong limitation is represented by the need for large high-fidelity datasets for training, which are generally not available for real-life structures. A possible solution is represented by unsupervised machine learning methods. Such methods have already been employed for damage detection in the SHM field, but no fully unsupervised approach to damage localization has been proposed yet. Hence, in this work, a framework combining an unsupervised data-driven method and an in-house tomographic approach is presented to process LWs to localize damage. Specifically, convolutional auto-associative neural networks (CAANNs) are employed to process diagnostic signals without requiring any prior feature extraction process. Furthermore, CAANNs predictions are processed to generate damage probability maps. The performance of the proposed method was tested against a numerical case study involving an Aluminum plate and against two experimental datasets of LWs acquired on a full-scale composite wing and a composite plate. Results showed that the proposed method outperformed classic damage diagnosis algorithms in terms of damage localization accuracy.

Examining urban delivery service user profiles and determinants of drone delivery adoption in Ghana considering usage before and after the COVID-19 pandemic

The demand for urban delivery services (UDS) in low- and middle-income countries (LMICs) has been growing due to the spike in e-commerce activities. This phenomenon has led to the search for more innovative UDS modes, such as drones, as they provide fast, safer, and eco-friendly services. Although researchers have examined the perceptions regarding the adoption of drone delivery, there are many unknowns about the intentions to use this futuristic delivery mode. By employing survey data obtained through an online survey conducted in Ghana, this study aims to gain a deeper understanding of the current UDS user profiles and investigate their intentions to adopt drone deliveries in the future, considering the lived experiences of traditional UDS usage pre/post the COVID-19 pandemic. Regarding UDS user profiling, the study applied an advanced unsupervised machine learning method known as cluster correspondence analysis to simultaneously identify critical clusters in the data as well as key factor associations for each group. The findings demonstrated that most people in Ghana who seldom used the traditional UDS usually experienced poor service levels. Females living in the capital of Ghana who are frequent users of UDS mainly experienced acceptable service levels. Using an ordered probit regression model, the study identified that variables such as males, people who desired contactless UDS, and frequent UDS users at work or school during the COVID-19 pandemic were more inclined to embrace drone deliveries when introduced. Conversely, individuals who frequently used UDS services at home, those living in large cities, and those with lower income levels expressed reluctance toward drone delivery services. This study augments the urban logistics literature in LMICs by offering insightful suggestions in the areas of education, enhanced marketing strategies, and delivery optimization, which are instrumental for informing policy decisions. Government-related transportation bodies, private logistics firms, and trade unions can collaborate and adopt the recommendations provided to improve urban delivery services and to pave the way for integrating drone delivery services in the near future.

Towards Benchmarking for Evaluating Machine Learning Methods in Detecting Outliers in Process Datasets

Within the integration and development of data-driven process models, the underlying process is digitally mapped in a model through sensory data acquisition and subsequent modelling. In this process, challenges of different types and degrees of severity arise in each modelling step, according to the Cross-Industry Standard Process for Data Mining (CRISP-DM). Particularly in the context of data acquisition and integration into the process model, it can be assumed with a sufficiently high degree of probability that the acquired data contain anomalies of various kinds. The outliers must be detected in the data preparation and processing phase and dealt with accordingly. If this is sufficiently implemented, it will positively impact the subsequent modelling in terms of accuracy and precision. Therefore, this paper shows how outliers can be identified using the unsupervised machine learning methods autoencoder, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Isolation Forest (iForest), and One-Class Support Vector Machine (OCSVM). Following implementing these methods, we compared them by applying the Numenta Anomaly Benchmark (NAB) and sufficiently presented the individual strengths and disadvantages. Evaluating the correctness, distinctiveness and robustness criteria described in the paper showed that the One-Class Support Vector Machine was outstanding among the methods considered. This is because the OCSVM achieved acceptable anomaly detections on the available process datasets with comparatively little effort.

Developing a Saturation Height Function (SHF) for the classified Hydraulic Flow Unit, a case study from the Middle Miocene Carbonate reservoir in the Song Hong Basin, Vietnam

The Middle Miocene carbonate is the main gas bearing reservoir in the southern part of the Song Hong basin. Carbonate reservoirs present challenges for engineers and geologists to characterize because of their complexity due to depositional and diagenetic processes. The log based calculated water saturation Sw for this complex carbonate reservoir bears a lot of uncertainties due to its heterogeneity that strongly affects to m and n values in the water saturation equation. As the calculated Sw strongly affects the Hydrocarbon Initially In Place (HIIP) estimation, the production performance prediction, etc., the alternative method using the Saturation Height Function (SHF) based on the capillary pressure was introduced. The SHF building process uses the combination of the result of HFU classification and permeability prediction given by machine learning methods, which have become a very useful tool for complex reservoir characterization such as carbonate reservoir for decades. Our obtained results allowed us to subdivide the carbonate reservoir into 5 Hydraulic Flow Unit (HFU) by using unsupervised machine learning methods, and then the capillary pressure (Pc) curves were classified and a proper saturation-height function using Skelt Harrison equation was assigned to individual HFU after testing with common SHF equations. The HFU number for each individual point was predicted based on the log data measured along the borehole using the supervised machine learning techniques while the Sw was calculated using the chosen SHF model and the FWL defined on log curves and pressure data from RCI measurement. The Sw computed from the SHF model is reliable and can be used for water saturation estimation for the whole field for known FWL and predicted HFU numbers.