Supervised Machine Learning Methods Research Articles

Supervised Parametric Learning in the Identification of Composite Biomarker Signatures of Type 1 Diabetes in Integrated Parallel Multi-Omics Datasets.

Type 1 diabetes (T1D) is a devastating autoimmune disease, and its rising prevalence in the United States and around the world presents a critical problem in public health. While some treatment options exist for patients already diagnosed, individuals considered at risk for developing T1D and who are still in the early stages of their disease pathogenesis without symptoms have no options for any preventive intervention. This is because of the uncertainty in determining their risk level and in predicting with high confidence who will progress, or not, to clinical diagnosis. Biomarkers that assess one's risk with high certainty could address this problem and will inform decisions on early intervention, especially in children where the burden of justifying treatment is high. Single omics approaches (e.g., genomics, proteomics, metabolomics, etc.) have been applied to identify T1D biomarkers based on specific disturbances in association with the disease. However, reliable early biomarkers of T1D have remained elusive to date. To overcome this, we previously showed that parallel multi-omics provides a more comprehensive picture of the disease-associated disturbances and facilitates the identification of candidate T1D biomarkers. This paper evaluated the use of machine learning (ML) using data augmentation and supervised ML methods for the purpose of improving the identification of salient patterns in the data and the ultimate extraction of novel biomarker candidates in integrated parallel multi-omics datasets from a limited number of samples. We also examined different stages of data integration (early, intermediate, and late) to assess at which stage supervised parametric models can learn under conditions of high dimensionality and variation in feature counts across different omics. In the late integration scheme, we employed a multi-view ensemble comprising individual parametric models trained over single omics to address the computational challenges posed by the high dimensionality and variation in feature counts across the different yet integrated multi-omics datasets. the multi-view ensemble improves the prediction of case vs. control and finds the most success in flagging a larger consistent set of associated features when compared with chance models, which may eventually be used downstream in identifying a novel composite biomarker signature of T1D risk. the current work demonstrates the utility of supervised ML in exploring integrated parallel multi-omics data in the ongoing quest for early T1D biomarkers, reinforcing the hope for identifying novel composite biomarker signatures of T1D risk via ML and ultimately informing early treatment decisions in the face of the escalating global incidence of this debilitating disease.

Multi-instance learning based lung nodule system for assessment of CT quality after small-field-of-view reconstruction

Small-field-of-view reconstruction CT images (sFOV-CT) increase the pixel density across airway structures and reduce partial volume effects. Multi-instance learning (MIL) is proposed as a weakly supervised machine learning method, which can automatically assess the image quality. The aim of this study was to evaluate the disparities between conventional CT (c-CT) and sFOV-CT images using a lung nodule system based on MIL and assessments from radiologists. 112 patients who underwent chest CT were retrospectively enrolled in this study between July 2021 to March 2022. After undergoing c-CT examinations, sFOV-CT images with small-field-of-view were reconstructed. Two radiologists analyzed all c-CT and sFOV-CT images, including features such as location, nodule type, size, CT values, and shape signs. Then, an MIL-based lung nodule system objectively analyzed the c-CT (c-MIL) and sFOV-CT (sFOV-MIL) to explore their differences. The signal-to-noise ratio of lungs (SNR-lung) and contrast-to-noise ratio of nodules (CNR-nodule) were calculated to evaluate the quality of CT images from another perspective. The subjective evaluation by radiologists showed that feature of minimal CT value (p = 0.019) had statistical significance between c-CT and sFOV-CT. However, most features (all with p < 0.05), except for nodule type, location, volume, mean CT value, and vacuole sign (p = 0.056–1.000), had statistical differences between c-MIL and sFOV-MIL by MIL system. The SNR-lung between c-CT and sFOV-CT had no statistical significance, while the CNR-nodule showed statistical difference (p = 0.007), and the CNR of sFOV-CT was higher than that of c-CT. In detecting the difference between c-CT and sFOV-CT, features extracted by the MIL system had more statistical differences than those evaluated by radiologists. The image quality of those two CT images was different, and the CNR-nodule of sFOV-CT was higher than that of c-CT.

Genomic prediction using machine learning: a comparison of the performance of regularized regression, ensemble, instance-based and deep learning methods on synthetic and empirical data

BackgroundThe accurate prediction of genomic breeding values is central to genomic selection in both plant and animal breeding studies. Genomic prediction involves the use of thousands of molecular markers spanning the entire genome and therefore requires methods able to efficiently handle high dimensional data. Not surprisingly, machine learning methods are becoming widely advocated for and used in genomic prediction studies. These methods encompass different groups of supervised and unsupervised learning methods. Although several studies have compared the predictive performances of individual methods, studies comparing the predictive performance of different groups of methods are rare. However, such studies are crucial for identifying (i) groups of methods with superior genomic predictive performance and assessing (ii) the merits and demerits of such groups of methods relative to each other and to the established classical methods. Here, we comparatively evaluate the genomic predictive performance and informally assess the computational cost of several groups of supervised machine learning methods, specifically, regularized regression methods, deep, ensemble and instance-based learning algorithms, using one simulated animal breeding dataset and three empirical maize breeding datasets obtained from a commercial breeding program.ResultsOur results show that the relative predictive performance and computational expense of the groups of machine learning methods depend upon both the data and target traits and that for classical regularized methods, increasing model complexity can incur huge computational costs but does not necessarily always improve predictive accuracy. Thus, despite their greater complexity and computational burden, neither the adaptive nor the group regularized methods clearly improved upon the results of their simple regularized counterparts. This rules out selection of one procedure among machine learning methods for routine use in genomic prediction. The results also show that, because of their competitive predictive performance, computational efficiency, simplicity and therefore relatively few tuning parameters, the classical linear mixed model and regularized regression methods are likely to remain strong contenders for genomic prediction.ConclusionsThe dependence of predictive performance and computational burden on target datasets and traits call for increasing investments in enhancing the computational efficiency of machine learning algorithms and computing resources.

Analyzing credit card fraud cases with supervised machine learning methods: logistic regression and Naive Bayes

Frauds involving credit cards are simple and effortless to target. With the rise of online payment credit cards have had a huge role in our daily life and economy for the past two decades and it is an important task for companies to identify fraud and non-fraud transactions. As the number of credit cards grows every day and the volume of transactions increases quickly in tandem, fraudsters who wish to exploit this market for illegitimate gains have come to light. Nowadays, it’s quite easy to access anyone’s credit card information, which makes it simpler for card fraudsters to do their crimes. Thanks to advances in technology, it is now possible to determine whether information gained with malicious intent has been used by looking at the costs and time involved in altering account transactions. The Credit Card Fraud analysis data set, which is obtained from the Kaggle database, is used in the modeling process together with The Logistic regression method and Naive Bayes algorithms. Using the Knime platform, we are going to apply machine learning techniques to practical data in this study. The goal of this study is to identify who performed the transaction by examining the periods when people use their credit cards. The Logistic regression approach and the Naive Bayes method both had success rates of 99.83%, which is the highest. The two methods’ results are based on Cohen’s kappa, accuracy, precision, recall, and other metrics.

Classification of Pepper Seeds by Machine Learning Using Color Filter Array Images.

The purpose of this work is to classify pepper seeds using color filter array (CFA) images. This study focused specifically on Penja pepper, which is found in the Litoral region of Cameroon and is a type of Piper nigrum. India and Brazil are the largest producers of this variety of pepper, although the production of Penja pepper is not as significant in terms of quantity compared to other major producers. However, it is still highly sought after and one of the most expensive types of pepper on the market. It can be difficult for humans to distinguish between different types of peppers based solely on the appearance of their seeds. To address this challenge, we collected 5618 samples of white and black Penja pepper and other varieties for classification using image processing and a supervised machine learning method. We extracted 18 attributes from the images and trained them in four different models. The most successful model was the support vector machine (SVM), which achieved an accuracy of 0.87, a precision of 0.874, a recall of 0.873, and an F1-score of 0.874.

ESG Factors and Firms' Credit Risk

We explore the relationship between credit risk and Environmental, Social, and Governance (ESG) dimensions using Supervised Machine Learning (SML) techniques on a cross-section of European listed companies. Our proxy for credit risk is the z-score originally proposed by Altman (1968). As potential explanatory variables, we consider an extensive number of raw ESG factors sourced from the rating provider MSCI. In the first stage, we demonstrate, using different SML methods such as LASSO and Random Forest, that a selection of ESG factors, in addition to the usual accounting ratios, helps explaining a firm’s probability of default. In the second stage, we measure the impact of the selected variables on the risk of default. Our approach provides a novel perspective to understand which ESG factors may be associated with the credit score of individual companies.

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.

Morphological Species Delimitation in The Western Pond Turtle (Actinemys): Can Machine Learning Methods Aid in Cryptic Species Identification?

As the discovery of cryptic species has increased in frequency, there has been an interest in whether geometric morphometric data can detect fine-scale patterns of variation that can be used to morphologically diagnose such species. We used a combination of geometric morphometric data and an ensemble of five supervised machine learning methods (MLMs) to investigate whether plastron shape can differentiate two putative cryptic turtle species, Actinemys marmorata and Actinemys pallida. Actinemys has been the focus of considerable research due to its biogeographic distribution and conservation status. Despite this work, reliable morphological diagnoses for its two species are still lacking. We validated our approach on two datasets, one consisting of eight morphologically disparate emydid species, the other consisting of two subspecies of Trachemys (T. scripta scripta, T. scripta elegans). The validation tests returned near-perfect classification rates, demonstrating that plastron shape is an effective means for distinguishing taxonomic groups of emydids via MLMs. In contrast, the same methods did not return high classification rates for a set of alternative phylogeographic and morphological binning schemes in Actinemys. All classification hypotheses performed poorly relative to the validation datasets and no single hypothesis was unequivocally supported for Actinemys. Two hypotheses had machine learning performance that was marginally better than our remaining hypotheses. In both cases, those hypotheses favored a two-species split between A. marmorata and A. pallida specimens, lending tentative morphological support to the hypothesis of two Actinemys species. However, the machine learning results also underscore that Actinemys as a whole has lower levels of plastral variation than other turtles within Emydidae, but the reason for this morphological conservatism is unclear.

Machine Learning Techniques for Crop Yield Forecasting in Semi-Arid (3A) Zone, Rajasthan (India)

Economic growth and prosperity of a nation are inextricably linked to the agricultural sector. In the compass of agriculture, climate and other environmental changes are one of the main challenges. The present study attempts to predict crop yield for the Jaipur district which is an important region in the semi-arid eastern plain of Rajasthan (India). Machine learning (ML) techniques are used in forecasting and developing practical solutions for numerous challenges such as climate change with other environmental factors. Crop yield prediction is the process of predicting yield using historical data through meteorological parameters and past yield records. This paper used the agrometeorological time-series data from the year 1991 to 2020 for optimal yield forecasting. There have been numerous attempts to improve crop yield prediction by employing machine learning techniques. However, in this study, fusing the intelligence of reinforcement with deep learning, we got a comprehensive framework for mapping raw data to crop prediction values, allowing an optimal estimation of crop yields with higher accuracy. Upon comparative analysis of numerous ML algorithms, Random Forest is found the best-performing algorithm with an accuracy of 92.3% using supervised machine learning methods. With an accuracy of 92.3%, the proposed Random Forest-based model outperforms other techniques that are currently being used to predict crop yields. The study predictions could significantly help in choosing the best cropping pattern and planning for action accordingly. The results provide the best ways to solve environmental and agricultural problems in this semi-arid region of the specified Rajasthan state facing climate change issues.

Heart Disease Detection Using Machine Learning Models

An enormous number of deaths occur every year as a result of heart disease, making it a major concern in world health. Improving patient outcomes and lowering death rates, early detection and correct diagnosis of cardiac disease play a key role. When the heart's arteries become blocked, oxygen-poor blood cannot reach the heart properly, resulting in coronary heart disease. Early detection of cardiac disease is viable because it reduces medical costs and potentially saves the patient's life. Recently presented methods have improved heart failure detection accuracy on testing data without sacrificing accuracy on training data, yet most of these algorithms are suffering from the issue of overfitting. Models that were created end up fitting the test data too well. In this study, we create a novel diagnostic system to address this issue, and the resulting system demonstrates high intelligence and excellent performance on both training and testing data. Machine learning (ML) algorithms have demonstrated promising potential in assisting healthcare professionals with timely and accurate diagnosis. In this paper, is based on supervised machine learning methods are decision tree (DT), random forest (RF), Support vector Machine (SVM), Principal Component Analysis(PCA). We compare their accuracy with each other by using bar plot.

Predicting Pandemic Fatality Based on Supervised Machine Learning Methods

Predicting Pandemic Fatality Based on Supervised Machine Learning Methods

Investigation of early detection possibilities of sugar beet disease with machine learning algorithms based on multispectral reflection

Traditionally, diagnosis and monitoring of agricultural diseases are carried out through on-site observation and inspection. These methods are time-consuming and may represent limited samples. Therefore, remote sensing technology has become an important tool in disease detection and monitoring in agriculture. In the research, Cercospora leaf spot (Cercospora beticola sacc.) disease, which cause significant economic losses in sugar beet production, were detected in the early stages using machine learning algorithms using non-invasive multispectral images taken with UAV under field conditions is intended to be determined. The research was fulfilled using images from the grower fields in two regions where the disease was observed intensively. Index value data from digital surface model maps created by processing the images taken were used as training and test data. Numerical data was tested using five different supervised machine learning methods. The success of the analyzed models in predicting disease formation from the index values obtained from the images taken and the physiological changes that occur before the disease agents appear on sugar beet leaves was over 70%. Among the models compared in the study, the k-nearest neighbor classifier (KNN) model gave the highest success in both diseases, with 83% accuracy and 76% and 86% f1-score values. The support vector machines model followed the KNN model with 77% accuracy, 75%, and 86% f1-score values. According to the results of the research, it has been revealed that plant diseases have the potential for pre-symptomatic detection, and by processing the images obtained with UAV-based MS images, it is possible to detect diseases in the early period.

Development and research of an intelligent control system and protection of electrical network equipment

The key element of an intelligent energy system with an active-adaptive network is relay protection, designed to ensure uninterrupted and stable operation of the energy system. The goal of the work is the development and research of intelligent control systems, protection and automation of electrical networks. Research has been carried out on an intelligent security system to determine the boundaries of restrictions in the use of artificial intelligence. The authors supplemented the existing complex with physical models of the power supply system with a digital twin, an artificial neural network, and studied the complex in order to answer some questions about the use of a neural network in relay protection. An iterative approach to software development, supervised machine learning method, and Python programming language were used. Pychram Community Edition was chosen as the integrated development environment. Intelligent system software has been developed, including a digital twin, an executive part and a neural network model, which makes it possible to recognize types of short circuits in the electrical network. The works of domestic and foreign authors on this topic are considered.

Predicting incident heart failure from population-based nationwide electronic health records: protocol for a model development and validation study

IntroductionHeart failure (HF) is increasingly common and associated with excess morbidity, mortality, and healthcare costs. Treatment of HF can alter the disease trajectory and reduce clinical events in HF. However,...

Machine-learning based risk prediction of in-hospital outcomes following STEMI: the STEMI-ML score.

Traditional prognostic models for ST-segment elevation myocardial infarction (STEMI) have limitations in statistical methods and usability. We aimed to develop a machine-learning (ML) based risk score to predict in-hospital mortality, intensive care unit (ICU) admission, and left ventricular ejection fraction less than 40% (LVEF < 40%) in STEMI patients. We reviewed 1,863 consecutive STEMI patients undergoing primary percutaneous coronary intervention (pPCI) or rescue PCI. Eight supervised ML methods [LASSO, ridge, elastic net (EN), decision tree, support vector machine, random forest, AdaBoost and gradient boosting] were trained and validated. A feature selection method was used to establish more informative and nonredundant variables, which were then considered in groups of 5/10/15/20/25/30(all). Final models were chosen to optimise area under the curve (AUC) score while ensuring interpretability. Overall, 128 (6.9%) patients died in hospital, with 292 (15.7%) patients requiring ICU admission and 373 (20.0%) patients with LVEF < 40%. The best-performing model with 5 included variables, EN, achieved an AUC of 0.79 for in-hospital mortality, 0.78 for ICU admission, and 0.74 for LVEF < 40%. The included variables were age, pre-hospital cardiac arrest, robust collateral recruitment (Rentrop grade 2 or 3), family history of coronary disease, initial systolic blood pressure, initial heart rate, hypercholesterolemia, culprit vessel, smoking status and TIMI flow pre-PCI. We developed a user-friendly web application for real-world use, yielding risk scores as a percentage. The STEMI-ML score effectively predicts in-hospital outcomes in STEMI patients and may assist with risk stratification and individualising patient management.

SSIOE: Self-Supervised Indoor Occupancy Estimation for Intelligent Building Management

This paper aims to estimate indoor occupancy given the real-time observed signals from the existing sensors; next, as a practical application, we build a dynamic control schedule for energy saving based on the estimated indoor occupancy. However, several issues need to be addressed. First, it is impossible to train the model with rich labels due to the expensive labeling cost. Second, manual annotation of the continuous occupancy rate is complex. Third, the mapping relationship between sensor data and occupancy will change in the long run. In this paper, we proposed a new algorithm named Self-Supervised Indoor Occupancy Estimation (SSIOE) to overcome the challenges. Specifically, our training scheme aims to (I) generate a set of pseudo labels in a simple way to mark the time periods believed to be either in a high or low occupancy state and (II) utilize these sparse labels for training a network to infer the continuous occupancy ratio. By reformulating the problem as a Wasserstein-distance-like estimation, SSIOE is a novel learning-based method that can rely only on the weak/sparse labels of either “high-occupancy” or “low-occupancy” and learn to estimate the continuous occupancy ratio. Furthermore, to deal with the scarce annotation problem, we proposed a novel physical constraint loss to model the physical prior. (III) Last but not least, to strengthen the adaptability, we integrated Model-Agnostic Meta-Learning (MAML) to train the model for dynamic model updating. Experimental results show that SSIOE can provide reliable occupancy estimation and flexibly adapt to various control modes without retraining the model. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This paper is motivated by the real-time occupancy estimation problem, which is crucial for building management systems to arrange lighting and air conditioning. A dynamic schedule adapted to the occupancy would help balance the user’s comfort and energy saving; however, installing the occupancy sensors raises additional costs. Therefore, this paper proposes SSIOE, a label-free machine learning algorithm that can learn to estimate the continuous occupancy ratio given the sensor status in real time. Prior works trained upon supervised machine learning methods suffer from heavy manual annotation workload, while the unsupervised machine learning approaches fail to provide continuous occupancy estimation values. This paper addresses the above challenges by proposing a novel self-supervised training scheme. Preliminary experiments suggest that this approach is feasible to work in real-world buildings. However, it should be noted that the occupancy estimated by SSIOE is a ratio (0.0-1.0), which is relative to the maximum and minimum occupancy of the training data (or fine-tuning data). For example, if the maximum number of occupants in the training data is 200 and the minimum number of occupants is 50, then the estimated occupancy value of 1.0 is about 200 occupants, and the occupancy value of 0.0 is about 50 occupants. When the maximum and minimum occupants of the space change, the model needs to be updated to capture the new mapping of the estimated occupancy ratio to the number of occupants. In future research, we plan to further improve and simplify the adaptation ability of SSIOE by continuously and automatically updating the system with the latest collected as the system runs.

Renewing Iterative Self-Labeling Domain Adaptation With Application to the Spine Motion Prediction

The area of transfer learning comprises supervised machine learning methods that cope with the issue when the training and testing data have different input feature spaces or distributions. In this work, we propose a novel transfer learning algorithm called Renewing Iterative Self-labeling Domain Adaptation (Re-ISDA). This work is motivated by a cervical spine motion prediction problem, with the goal of predicting the cervical spine motion of different subjects using the measurements of “exterior features”. The joint distribution of the exterior features for each subject may vary with one’s distinct BMI, sex and other characteristics; the sample size of this problem is limited due to the high experimental cost. The proposed method is well suited for transfer learning problems with limited training samples. The learning problem is formulated as a dynamic programming model and the latter is then solved by an efficient greedy algorithm. Numerical studies show that the proposed method outperforms prevailing transfer learning methods. The proposed method also achieves high prediction accuracy for the cervical spine motion problem. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This paper is motivated by the problem of using human body exterior features (e.g., contours, surface markers) acquired by computer vision or other measurement means to determine the underlying physics or mechanics. The goal is to build a subject-specific model to predict cervical spine motion based on surface-based measurements. Conventional supervised learning and transfer learning methods fail to give an accurate prediction due to the subject-to-subject variation and/or the small sample size of the training data. This paper proposes a novel transfer learning method called renewing iterative self-labeling domain adaptation to modify the possible error of the pseudo labels and thus make the training process more stable. The proposed method improves the accuracy by over <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$35\%$</tex-math> </inline-formula> on the traditional transfer learning methods. This method is applicable to limited-sample problems with different training and testing input distributions, and is particularly suitable for time-series data.

Pure Data-Driven Machine Learning Challenges for pFMEA: A Case Study

Manufacturing processes are susceptible to quality defects, resulting in overall equipment effectiveness reduction. Proactive and reactive methods, such as process failure mode and effects analysis, and root cause analysis, have been developed to eliminate potential causes of failure modes. In this study, data from an assembly case is evaluated using supervised machine learning methods to analyze the challenges of purely data-driven failure mode detection. Assembly step execution times, as indicators, and end-of-the-line quality checklists, as the failure modes, are used to gain insights into failure mode detection. Challenges for data-driven methods are discussed and possible future research streams are proposed.

Automated Detection of Heart Attacks: A CNN-Based Approach

Abstract: Seventeen million people die because of the world’s cardiovascular diseases annually that make up to 32% of the global population. One fourth (25%) of global deaths and 17 million each year, representing almost a third (32%), are attributable to cardiovascular disease or the world heart attack syndrome. In addition, cardiovascular disease kills an amazing 32% of overall deaths on the globe per quarter of percentage to be exact. In the short term, heart disease’s mortality rate could decline by predicting it. It has demonstrated that convolutional neural network is the best approach when implementing deep learning, generating training and test accuracy of 94.2%. Deep learning algorithm comprises of artificial neuronal network that uses a supervised machine learning method for the data set prediction. The EGC can also be used to detect cardio diseases. Deep learning will turn out being the most preferred approach because of its high precision and predictive power which also translates to good sensitivity and specificity. Deep learning techniques give better results than others, for example, CNN with autoencoders.

Predicting financial distress using supervised machine learning algorithms: an application on Borsa Istanbul

Purpose- The main purpose of this study is to identify the most significant variables to detect financial distress earlier and to find the best machine learning algorithm model. Methodology-This study has used Support Vector Machine, Logistic Regression, Random Forest and K-nearest neighbors method techniques to predict the financial distress prediction for the companies of Turkey between 2012 and 2021. Findings- As a result of the study, it has been determined that Random Forest provides the best results in terms of precision, accuracy, and recall. Further, this study has found the most important five independent variables to determine the financial distress status of the firms. In this way, it has been found that Current Assets/ Current Liabilities, Working Capital / Total Assets, Gross profit / Revenue, Retained Earnings / Total Assets and Sales growth rate are the most useful variables to determine financial distress status of Turkish firms earlier. Conclusion- This study has concluded that cash ratios and profitability ratios and sales growth are the most important independent variables to determine financial distress one-year ahead. Furthermore, it has been found that random forest is the best machine learning method among other supervised machine learning methods used in this study. Keywords: Financial distress, support vector machine, logistic regression, random forest, k-nearest neighbors JEL Codes: G32, G33, C52