Binary Classification Problem Research Articles

Refactoring and performance analysis of the main CNN architectures: using false negative rate minimization to solve the clinical images melanoma detection problem

BackgroundMelanoma is one of the deadliest tumors in the world. Early detection is critical for first-line therapy in this tumor pathology and it remains challenging due to the need for histological analysis to ensure correctness in diagnosis. Therefore, multiple computer-aided diagnosis (CAD) systems working on melanoma images were proposed to mitigate the need of a biopsy. However, although the high global accuracy is declared in literature results, the CAD systems for the health fields must focus on the lowest false negative rate (FNR) possible to qualify as a diagnosis support system. The final goal must be to avoid classification type 2 errors to prevent life-threatening situations. Another goal could be to create an easy-to-use system for both physicians and patients.ResultsTo achieve the minimization of type 2 error, we performed a wide exploratory analysis of the principal convolutional neural network (CNN) architectures published for the multiple image classification problem; we adapted these networks to the melanoma clinical image binary classification problem (MCIBCP). We collected and analyzed performance data to identify the best CNN architecture, in terms of FNR, usable for solving the MCIBCP problem. Then, to provide a starting point for an easy-to-use CAD system, we used a clinical image dataset (MED-NODE) because clinical images are easier to access: they can be taken by a smartphone or other hand-size devices. Despite the lower resolution than dermoscopic images, the results in the literature would suggest that it would be possible to achieve high classification performance by using clinical images. In this work, we used MED-NODE, which consists of 170 clinical images (70 images of melanoma and 100 images of naevi). We optimized the following CNNs for the MCIBCP problem: Alexnet, DenseNet, GoogleNet Inception V3, GoogleNet, MobileNet, ShuffleNet, SqueezeNet, and VGG16.ConclusionsThe results suggest that a CNN built on the VGG or AlexNet structure can ensure the lowest FNR (0.07) and (0.13), respectively. In both cases, discrete global performance is ensured: 73% (accuracy), 82% (sensitivity) and 59% (specificity) for VGG; 89% (accuracy), 87% (sensitivity) and 90% (specificity) for AlexNet.

Identification of Key Gene Modules and Novel Transcription Factors in Tetralogy of Fallot Using Machine Learning and Network Topological Features

Tetralogy of Fallot (TOF) is a combinatorial congenital abnormality comprising of ventricular septal defect (VSD), pulmonary valve stenosis, a misplaced aorta and a thickened right ventricular wall. Biologically relevant module identification from transcriptome data may be considered as a binary classification problem. We utilized publicly accessible mRNA expression data to extract the differentially expressed genes (DEGs) and further weighted gene co-expression network analysis to identify ten modules in TOF. Network topological properties of modular and non-modular genes were considered as features for binary classification. We applied SVM, Random Forest, Decision Trees, KNN and Naïve Bayes algorithm to network features. Random Forest and decision tree algorithms displayed an accuracy of 99.1% and 98% respectively. All the methods, in combination predicted 71 common genes which were used to construct a gene regulatory network. The network was expanded to include 30 miRNAs targeting the genes. Interestingly, 39 out of 71 genes were transcription factors out of which ELN, SOX6 and FOXO3 genes are novel candidates in TOF. The work also provides a sub-module of genes and miRNAs supported by statistical models as prospective candidates to be biomarkers.

A decision-making tool for the determination of the distribution center location in a humanitarian logistics network

The distribution of humanitarian aid is a vital issue for humanity's future. In recent years, the management of humanitarian crises has become more crucial than it was a decade ago. Due to the volatility and urgency that characterize such situations, one of the most important challenges globally is the optimization of decisions regarding the timely distribution of aid during humanitarian operations. Our main goal is to develop an innovative decision-making tool, essential for non-profit organizations and governments that aims at the prompt selection of the location of the distribution center of humanitarian aid, in cases of natural or human-made disasters. The proposed tool is based on network science principles and can be used for selecting a suitable node for the installation of a distribution center during the beginning of a humanitarian crisis, considering that networks have a volatile nature and require quick decisions. For the configuration of the proposed tool we use a combination of a classical heuristic algorithm and predictive models based on a binary classification problem with the support of a supervised deep neural network. It is developed using the R programming language with the contribution of the “Shiny” package (web application framework for R) along with other packages for network analysis, data manipulation and visualization.

Oversampling method based on GAN for tabular binary classification problems

Data-imbalanced problems are present in many applications. A big gap in the number of samples in different classes induces classifiers to skew to the majority class and thus diminish the performance of learning and quality of obtained results. Most data level imbalanced learning approaches generate new samples only using the information associated with the minority samples through linearly generating or data distribution fitting. Different from these algorithms, we propose a novel oversampling method based on generative adversarial networks (GANs), named OS-GAN. In this method, GAN is assigned to learn the distribution characteristics of the minority class from some selected majority samples but not random noise. As a result, samples released by the trained generator carry information of both majority and minority classes. Furthermore, the central regularization makes the distribution of all synthetic samples not restricted to the domain of the minority class, which can improve the generalization of learning models or algorithms. Experimental results reported on 14 datasets and one high-dimensional dataset show that OS-GAN outperforms 14 commonly used resampling techniques in terms of G-mean, accuracy and F1-score.

Predicting obsessive-compulsive disorder episodes in adolescents using a wearable biosensor-A wrist angel feasibility study.

Obsessive-compulsive disorders (OCD) are marked by distress, negative emotions, mental processes and behaviors that are reflected in physiological signals such as heart rate, electrodermal activity, and skin temperature. Continuous monitoring of physiological signals associated with OCD symptoms may make measures of OCD more objective and facilitate close monitoring of prodromal symptoms, treatment progress and risk of relapse. Thus, we explored the feasibility of capturing OCD events in the real world using an unobtrusive wrist worn biosensor and machine learning models. Nine adolescents (ages 10-17 years) with mild to moderate-severe OCD were recruited from child and adolescent mental health services. Participants were asked to wear the biosensor in the lab during conditions of rest and exposure to OCD symptom-triggering stimuli and for up to 8 weeks in their everyday lives and register OCD events. We explored the relationships among physiological data, registered OCD events, age, OCD symptom severity and symptom types. In the machine learning models, we considered detection of OCD events as a binary classification problem. A nested cross-validation strategy with either random 10-folds, leave-one-subject-out, or leave-week(s)-out in both layers was used. We compared the performance of four models: logistic regression, random forest (RF), feedforward neural networks, and mixed-effect random forest (MERF). To explore the ability of the models to detect OCD events in new patients, we assessed the performance of participant-based generalized models. To explore the ability of models to detect OCD events in future, unseen data from the same patients, we compared the performance of temporal generalized models trained on multiple patients with personalized models trained on single patients. Eight of the nine participants collected biosensor signals totaling 2, 405 h and registered 1, 639 OCD events. Better performance was obtained when generalizing across time compared to across patients. Generalized temporal models trained on multiple patients were found to perform better than personalized models trained on single patients. RF and MERF models outperformed the other models in terms of accuracy in all cross-validation strategies, reaching 70% accuracy in random and participant cross-validation. Our pilot results suggest that it is possible to detect OCD episodes in the everyday lives of adolescents using physiological signals captured with a wearable biosensor. Large scale studies are needed to train and test models capable of detecting and predicting episodes. ClinicalTrials.gov: NCT05064527, registered October 1, 2021.

Utilizing logistic regression to compare risk factors in disease modeling with imbalanced data: a case study in vitamin D and cancer incidence.

Imbalanced data, a common challenge encountered in statistical analyses of clinical trial datasets and disease modeling, refers to the scenario where one class significantly outnumbers the other in a binary classification problem. This imbalance can lead to biased model performance, favoring the majority class, and affecting the understanding of the relative importance of predictive variables. Despite its prevalence, the existing literature lacks comprehensive studies that elucidate methodologies to handle imbalanced data effectively. In this study, we discuss the binary logistic model and its limitations when dealing with imbalanced data, as model performance tends to be biased towards the majority class. We propose a novel approach to addressing imbalanced data and apply it to publicly available data from the VITAL trial, a large-scale clinical trial that examines the effects of vitamin D and Omega-3 fatty acid to investigate the relationship between vitamin D and cancer incidence in sub-populations based on race/ethnicity and demographic factors such as body mass index (BMI), age, and sex. Our results demonstrate a significant improvement in model performance after our undersampling method is applied to the data set with respect to cancer incidence prediction. Both epidemiological and laboratory studies have suggested that vitamin D may lower the occurrence and death rate of cancer, but inconsistent and conflicting findings have been reported due to the difficulty of conducting large-scale clinical trials. We also utilize logistic regression within each ethnic sub-population to determine the impact of demographic factors on cancer incidence, with a particular focus on the role of vitamin D. This study provides a framework for using classification models to understand relative variable importance when dealing with imbalanced data.

An empirical comparison of correlation-based systemic risk measures

Despite the growing attention in the last years on the topic of systemic risk, a widely accepted definition of systemic crisis is missing. We use a theoretical scheme to subjectively define a systemic event. This permits the analysis of a financial crisis as a standard binary classification problem, providing an intuitive and useful framework to compare systemic risk measures defined in very different fields. Then we focus the empirical analysis on the comparison of the performance of correlation-based systemic risk measures using the standard tools for the evaluation of binary classifiers as the receiver operating characteristic (ROC) curve and the area under the curve (AUC). We show that the binary classification framework is useful but unable to capture some significant differences among the measures under comparison. The experimental approach, developed on real financial data, is divided in an in-sample exercise, able to evaluate the descriptive power of the different systemic risk measures, and an out-of-sample application to evaluate the capacity of the measures in preventing and predicting systemic events. The forecasting ability of a measure can be fundamental for policy makers and investors respectively to stabilize market fluctuations and to reduce the losses.

An Educational Approach to Higgs Boson Hunting Using Machine Learning Classification Algorithms on ATLAS Open Data

In this study, the performance of several classification algorithms that are used to separate the H → ττ signal from background is investigated. The data set came from the publicly available ATLAS data, which was utilized for the Machine Learning (ML) competition. The data was obtained from a full ATLAS simulation and originated from proton-proton collisions. There are 250 thousand events in the data set, and 70% of them were used to train the algorithms. The primary objective of this research is to identify the signal events from the background events by using various ML methods in the context of high-energy physics. In order to discover a solution to the binary classification problem that was discussed earlier, six distinct classification algorithms were utilized. This article also compares the performance of these classification algorithms, including Linear Support Vector Machines (SVM), Radical SVM, Logistic Regression, K-Nearest Neighbours, XGBoost Classifier, and the AdaBoost Classifier. The best results were obtained using the XGBoost Classification method, which had an AUC of 0.84 ± 1.9 x 10-3 followed by the AdaBoost Classifier with an AUC of 0.82 ± 2.5 x 10-3.

Neural Network Based Approach to Recognition of Meteor Tracks in the Mini-EUSO Telescope Data

Mini-EUSO is a wide-angle fluorescence telescope that registers ultraviolet (UV) radiation in the nocturnal atmosphere of Earth from the International Space Station. Meteors are among multiple phenomena that manifest themselves not only in the visible range but also in the UV. We present two simple artificial neural networks that allow for recognizing meteor signals in the Mini-EUSO data with high accuracy in terms of a binary classification problem. We expect that similar architectures can be effectively used for signal recognition in other fluorescence telescopes, regardless of the nature of the signal. Due to their simplicity, the networks can be implemented in onboard electronics of future orbital or balloon experiments.

A robust twin support vector machine based on fuzzy systems

PurposeTwin support vector machine (TSVM) is an effective machine learning technique. However, the TSVM model does not consider the influence of different data samples on the optimal hyperplane, which results in its sensitivity to noise. To solve this problem, this study proposes a twin support vector machine model based on fuzzy systems (FSTSVM).Design/methodology/approachThis study designs an effective fuzzy membership assignment strategy based on fuzzy systems. It describes the relationship between the three inputs and the fuzzy membership of the sample by defining fuzzy inference rules and then exports the fuzzy membership of the sample. Combining this strategy with TSVM, the FSTSVM is proposed. Moreover, to speed up the model training, this study employs a coordinate descent strategy with shrinking by active set. To evaluate the performance of FSTSVM, this study conducts experiments designed on artificial data sets and UCI data sets.FindingsThe experimental results affirm the effectiveness of FSTSVM in addressing binary classification problems with noise, demonstrating its superior robustness and generalization performance compared to existing learning models. This can be attributed to the proposed fuzzy membership assignment strategy based on fuzzy systems, which effectively mitigates the adverse effects of noise.Originality/valueThis study designs a fuzzy membership assignment strategy based on fuzzy systems that effectively reduces the negative impact caused by noise and then proposes the noise-robust FSTSVM model. Moreover, the model employs a coordinate descent strategy with shrinking by active set to accelerate the training speed of the model.

A Novel Learning-Based PLST Algorithm for Multi-Label Classification

In Multi-Label Classification (MLC), each data sample is characterized by multiple labels. In ML, there is no restriction on the number of classes a data sample could belong to. Various methods are proposed to construct a model to map data samples to the class labels. Transforming the MLC problem into several binary classification problems is the baseline approach, also called the binary relevance technique. As an alternative solution, the label powerset approach transforms the MLC problem into a multi-class classification problem by building a binary classifier for every label composition. However, in Ensemble Methods (EMs), multiple classifiers are consolidated to form a multi-label ensemble classifier. In this case, each classifier predicts a class label. Thereupon, the predicted class labels are joined using an ensemble method. This paper presents a neoteric EM classifier using the Ridge Regression-based Principal Label Space Transformation (RR-PLST). In the proposed model, the PLST utilizes SVD to discover correlations between the labels and data samples. Dimensionality reduction is a well-known technique to enhance the model's precision and diminish execution time. Consequently, the OPLS technique is recruited to reduce data dimensions in the proposed model. Moreover, in the proposed model, a Q-Learning method is used to amend the initially predicted labels. To demonstrate the effectiveness of the proposed model, vast experiments are conducted on the various widely used multi-label datasets. The proposed algorithm results are compared with similar algorithms by the AP, EM, HS, Macro, and Micro F1 metrics. The experimental results demonstrate the superiority of the proposed model.

An adaptive error-correcting output codes algorithm based on gene expression programming and similarity measurement matrix

The multi-class classification task is one of the most common tasks in machine learning. As a typical solution based on a partitioning strategy, Error-Correcting Output Codes (ECOC) can transform a multi-class classification problem into multiple binary classification problems. The key of ECOC is to construct an effective codematrix to represent a set of class decomposition schemes, which transforms a multiclass problem into a group of binary class problems. Consequently, the design of a fast and effective ECOC codematrix generation method is of great research significance and value for solving multi-class classification problems. In ECOC algorithms, the design of codematrix is treated as a combination problem between different code columns, in which the evolutionary algorithm shows a great advantage. Based on this consideration, the Gene Expression Programming (GEP) is applied to search for the codematrix with high performance because its expressive tree structure makes it well represent codematrcies for subsequent optimization operations. This paper proposes an adaptive ECOC algorithm based on Gene Expression Programming (GEP) and similarity measurement matrix, named GEP-ECOC. In our GEP, each individual represents a set of columns to form a random ECOC codematrix, which is optimized in the evolutionary process. Meanwhile, the crossover and mutation operations are modified to include a legality checking process to ensure that the generated codematrix satisfies the ECOC constraints. The GEP-based ECOC codematrix generation algorithm can quickly produce a codematrix with better performance, which ensures the efficiency of the algorithm to a certain extent. In addition, an adaptive algorithm based on a similarity measurement matrix is proposed to add new columns to the current codematrix, aiming to better handle hard classes. Our algorithm is compared with other algorithms on various data sets, and the experimental results confirm that our GEP-ECOC can balance the efficiency and performance of the algorithm and achieve higher performance.

Headline-Driven Classification and Local Interpretation for Market Outperformance and Low-Risk Stock Prediction

Financial news is one of the most influential sources for making long-term investment decisions. The goal of this paper is to learn whether it is possible, based on texts from news headlines, to select stocks with low forecasted volatility which outrun the market represented by the Standard and Poor’s 500 Index. We solve several binary classification problems using a range of machine learning and deep learning principles. The best classifiers are interpreted with a model-agnostic technique, LIME, to extract key words having the greatest impact on the probabilities of market outrun and low volatility.

Self-adaptive micro-hole breakout detection in the electrochemical discharge drilling process based on CNN-BiLSTM

Electrochemical discharge machining (ECDM) is an effective and promising technology for the micro-hole drilling of hard and brittle insulating materials such as glass, quartz and ceramics. Hole exit quality is an important part of through-hole machining quality in ECDM. The hole exit quality is influenced by two different machining states before and after the hole breakout. Therefore, it is necessary to identify hole breakout occurrence to provide a basis for achieving a machining strategy that matches the changing machining state and thus improves the hole exit quality. However, currently, there is no effective method to detect the hole breakout in the ECDM of micro-holes. In this paper, a self-adaptive micro-hole breakout detection method based on CNN- BiLSTM was proposed. The hole breakout detection can be expressed as a binary classification problem according to the analysis of the current signals collected before and after the hole breakout. The sufficiency of the raw current signal information was discussed. A CNN-BiLSTM model was established to detect hole breakout using the current signals as the training dataset. The CNN-BiLSTM model showed better performance compared to CNN, LSTM, BiLSTM and CNN-LSTM by comprehensive considering of accuracy, F1 score, AUC value, training time, predict time, breakout detection results and Dunn test results. Furthermore, the length of current signals used for each single detection and the threshold value used for breakout detection were discussed. An optimal threshold value was calculated, and the delay of detection was within 66 ms. The proposed method was successfully applied to the hole breakout detection in the through-hole drilling experiments and the success rate was 100 %. The hole exit area just after breakout were measured and compared to the hole exit area after hole completion, the average ratios of the former to the latter were 11.69 %, 27.29 %, 38.43 % and 44.41 % for the voltage of 39 V, 41 V, 43 V and 45 V respectively.

Refined evaluation methods for preventive maintenance of project-level asphalt pavement based on confusion-regression model

Accurate evaluation of the performance of airport pavement can help managers make scientific and systematic preventive maintenance (PM) decisions, ensuring pavement safety and extending service life. To this end, firstly, this paper sorted out the pavement performance evaluation index system, based on which a multi-dimensional evaluation system for PM was proposed, including control indexes, macroscopic indexes, and microscopic indexes. Then, the pavement decay model based on the pavement condition index (PCI) was developed, and the regression model between PCI and International roughness index (IRI) and friction coefficient (μ) was established. Finally, by converting the PM decisions for project-level pavement into a binary classification problem in machine learning, the optimal maintenance thresholds and intervals for PM of each index were determined using receiver operating characteristic (ROC) curve and Kolmogorov-Smirnov (K-S) curve. The optimal threshold represents the threshold when the index can separate the binary classification problem (PM decisions) with maximum probability. Based on confusion-regression model, the pavement performance can be evaluated more comprehensively, and the timing of PM can be accurately determined. The model helps airport managers develop accurate PM plans to address future maintenance needs.

Machine Learning to Predict Faricimab Treatment Outcome in Neovascular Age-Related Macular Degeneration

To develop machine learning (ML) models to predict, at baseline, treatment outcomes at month 9 in patients with neovascular age-related macular degeneration (nAMD) receiving faricimab. Retrospective proof of concept study. Patients enrolled in the phase II AVENUE trial (NCT02484690) of faricimab in nAMD. Baseline characteristics and spectral domain-OCT (SD-OCT) image data from 185 faricimab-treated eyes were split into 80% training and 20% test sets at the patient level. Input variables were baseline age, sex, best-corrected visual acuity (BCVA), central subfield thickness (CST), low luminance deficit, treatment arm, and SD-OCT images. A regression problem (BCVA) and a binary classification problem (reduction of CST by 35%) were considered. Overall, 10 models were developed and tested for each problem. Benchmark classical ML models (linear, random forest, extreme gradient boosting) were trained on baseline characteristics; benchmark deep neural networks (DNNs) were trained on baseline SD-OCT B-scans. Baseline characteristics and SD-OCT data were merged using 2 approaches: model stacking (using DNN prediction as an input feature for classical ML models) and model averaging (which averaged predictions from the DNN using SD-OCT volume and from classical ML models using baseline characteristics). Treatment outcomes were defined by 2 target variables: functional (BCVA letter score) and anatomical (percent decrease in CST from baseline) outcomes at month9. The best-performing BCVA regression model with respect to the test coefficient of determination (R2) was the linear model in the model-stacking approach with R2 of 0.31. The best-performing CST classification model with respect to test area under receiver operating characteristics (AUROC) was the benchmark linear model with AUROC of 0.87. A post hoc analysis showed the baseline BCVA and the baseline CST had the most effect in the all-model prediction for BCVA regression and CST classification, respectively. Promising signals for predicting treatment outcomes from baseline characteristics were detected; however, the predictive benefit of baseline images was unclear in this proof-of-concept study. Further testing and validation with larger, independent datasets is required to fully explore the predictive capacity of ML models using baseline imaging data. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Improved prediction of MHC-peptide binding using protein language models.

Major histocompatibility complex Class I (MHC-I) molecules bind to peptides derived from intracellular antigens and present them on the surface of cells, allowing the immune system (T cells) to detect them. Elucidating the process of this presentation is essential for regulation and potential manipulation of the cellular immune system. Predicting whether a given peptide binds to an MHC molecule is an important step in the above process and has motivated the introduction of many computational approaches to address this problem. NetMHCPan, a pan-specific model for predicting binding of peptides to any MHC molecule, is one of the most widely used methods which focuses on solving this binary classification problem using shallow neural networks. The recent successful results of Deep Learning (DL) methods, especially Natural Language Processing (NLP-based) pretrained models in various applications, including protein structure determination, motivated us to explore their use in this problem. Specifically, we consider the application of deep learning models pretrained on large datasets of protein sequences to predict MHC Class I-peptide binding. Using the standard performance metrics in this area, and the same training and test sets, we show that our models outperform NetMHCpan4.1, currently considered as the-state-of-the-art.

Link Prediction for Temporal Heterogeneous Networks Based on the Information Lifecycle

Link prediction for temporal heterogeneous networks is an important task in the field of network science, and it has a wide range of real-world applications. Traditional link prediction methods are mainly based on static homogeneous networks, which do not distinguish between different types of nodes in the real world and do not account for network structure evolution over time. To address these issues, in this paper, we study the link prediction problem in temporal heterogeneous networks and propose a link prediction method for temporal heterogeneous networks (LP-THN) based on the information lifecycle, which is an end-to-end encoder–decoder structure. The information lifecycle accounts for the active, decay and stable states of edges. Specifically, we first introduce the meta-path augmented residual information matrix to preserve the structure evolution mechanism and semantics in HINs, using it as input to the encoder to obtain a low-dimensional embedding representation of the nodes. Finally, the link prediction problem is considered a binary classification problem, and the decoder is utilized for link prediction. Our prediction process accounts for both network structure and semantic changes using meta-path augmented residual information matrix perturbations. Our experiments demonstrate that LP-THN outperforms other baselines in both prediction effectiveness and prediction efficiency.

A New Deep Learning Framework for Imbalance Detection of a Rotating Shaft.

Rotor unbalance is the most common cause of vibration in industrial machines. The unbalance can result in efficiency losses and decreased lifetime of bearings and other components, leading to system failure and significant safety risk. Many complex analytical techniques and specific classifiers algorithms have been developed to study rotor imbalance. The classifier algorithms, though simple to use, lack the flexibility to be used efficiently for both low and high numbers of classes. Therefore, a robust multiclass prediction algorithm is needed to efficiently classify the rotor imbalance problem during runtime and avoid the problem's escalation to failure. In this work, a new deep learning (DL) algorithm was developed for detecting the unbalance of a rotating shaft for both binary and multiclass identification. The model was developed by utilizing the depth and efficacy of ResNet and the feature extraction property of Convolutional Neural Network (CNN). The new algorithm outperforms both ResNet and CNN. Accelerometer data collected by a vibration sensor were used to train the algorithm. This time series data were preprocessed to extract important vibration signatures such as Fast Fourier Transform (FFT) and Short-Time Fourier Transform (STFT). STFT, being a feature-rich characteristic, performs better on our model. Two types of analyses were carried out: (i) balanced vs. unbalanced case detection (two output classes) and (ii) the level of unbalance detection (five output classes). The developed model gave a testing accuracy of 99.23% for the two-class classification and 95.15% for the multilevel unbalance classification. The results suggest that the proposed deep learning framework is robust for both binary and multiclass classification problems. This study provides a robust framework for detecting shaft unbalance of rotating machinery and can serve as a real-time fault detection mechanism in industrial applications.

Analysis of COVID-19 cases and comorbidities using machine learning algorithms: A case study of the Limpopo Province, South Africa

This study examined the biological, social, and clinical risk factors for mortality in coronavirus of the year 2019 (COVID-19) hospitalised patients. The population of the study is prone to COVID-19, thus understanding the most common traits and comorbidities of people who were affected is crucial in reducing its consequences. In this study, four supervised machine learning algorithms were implemented and compared to predict the mortality rate based on the explanatory variables across the five districts of Limpopo Province in South Africa. The data was obtained from Limpopo Department of Health. Prediction about the chances of dying from COVID-19 disease was made using logistic regression, random forest, support vector machine, and decision tree algorithms on the dataset of 20,592 records with twenty-one attributes. Due to the imbalanced nature of the data, Random Over-Sampling Examples (ROSE) were employed to balance our data for more accurate classification effectively. The ROSE package provides functions to deal with binary classification problems in the presence of imbalanced classes. We used 70% of the data for training, while 30% was selected for testing the predictive algorithms. A technique called Step Akaike's Information Criterion (StepAIC) was deployed to reduce the insignificant variables from the full model of the logistic regression. According to the findings of the study, among the four algorithms tested, random forest had the highest recall rate for predicting mortality at roughly 79 percent compared to the other three algorithms. Accordingly, we conclude that random forest algorithm is appropriate for predicting the chances of patients dying from COVID-19 based on the attributes of the five districts of Limpopo Province. In terms of the features and their importance, a function called Variable Importance (VarImp) was used to check which of the attributes have predictive power on the outcome variable (discharged status). The findings revealed that age, ever ventilated, ever oxygenated, intensive ward upon admission, Waterberg district, and private facility type were among the risk factors that could be selected for the logistic regression model to predict mortality in hospitalised patients. This implies that special attention should be given to these identified variables. The random forest model is adequate to establish these factors since the findings reveal that a fairly considerable percentage of explained variation would correctly classify 79% of the cases. The novelty of conducting research on the analysis of COVID-19 cases and comorbidities using machine learning algorithms lies in the potential to uncover new insights and patterns that might not be immediately apparent through traditional statistical methods. Machine learning algorithms can quickly and accurately analyze large datasets and identify complex relationships between variables, which could provide valuable information for public health officials and policymakers.