Developing a decision support tool to predict delayed discharge from hospitals using machine learning

This study employs machine learning models to explore stock price prediction for Tenaga Nasional Berhad (TNB), Malaysia’s primary electricity provider. It addresses the limitations of previous studies by incorporating various input variables, including the stock market, technical, financial, and economic data. This study also tackles the issue of imbalanced class distribution due to small datasets of stock market data by generating synthetic data using Synthetic Minority Over-Sampling Technique (SMOTE) and Generative Adversarial Network-Synthetic Minority Over-Sampling Technique (GAN-SMOTE) techniques. The performance of four classifier models (random forest, Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), and Artificial Neural Network (ANN)) is evaluated without any synthetic data and with synthetic data generated. The SMOTE-ANN model is the best-performing model, exhibiting superior accuracy of 93%, F1-Score of 92%, precision of 90%, recall of 94%, and specificity of 92%. Overall, this research provides valuable insights into TNB stock price movements, offers a solution for imbalanced class distribution, and identifies the top-performing model for predicting TNB stock price movement. These findings are relevant to investors, analysts, and organisations in the utility sector.

Imbalanced data is a challenge for classification models. It reduces the overall performance of traditional learning algorithms. Besides, the minority class of imbalanced datasets is misclassified with a high ratio even though this is a crucial object of the classification process. In this paper, a new model called the Lasso-Logistic ensemble is proposed to deal with imbalanced data by utilizing two popular techniques, random over-sampling and random under-sampling. The model was applied to two real imbalanced credit data sets. The results show that the Lasso-Logistic ensemble model offers better performance than the single traditional methods, such as random over-sampling, random under-sampling, Synthetic Minority Oversampling Technique (SMOTE), and cost-sensitive learning.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium provided the original work is properly cited.

The classification of clinical notes into specific diagnostic categories is critical in healthcare, especially for mental health conditions like anxiety and adjustment disorder. In this study, we compare the performance of various artificial intelligence models, including both traditional machine learning approaches (random forest, support vector machine, K-nearest neighbors, decision tree, and eXtreme Gradient Boost) and deep learning models (DistilBERT and SciBERT), to classify clinical notes into these two diagnoses. Additionally, we implemented three oversampling strategies: No Oversampling, Random Oversampling, and Synthetic Minority Over-sampling Technique (SMOTE), to assess their impact on model performance. Hyperparameter tuning was also applied to optimize model accuracy. Our results indicate that oversampling techniques had minimal impact on model performance overall. The only exception was SMOTE, which showed a positive effect specifically with Bidirectional Encoder Representations from Transformers (BERT)-based models. However, hyperparameter optimization significantly improved accuracy across the models, enhancing their ability to generalize and perform on the dataset. The decision tree and eXtreme Gradient Boost models achieved the highest accuracy among machine learning approaches, both reaching 96%, while the DistilBERT and SciBERT models also attained 96% accuracy in the deep learning category. These findings underscore the importance of hyperparameter tuning in maximizing model performance. This study contributes to the ongoing research on AI-assisted diagnostic tools in mental health by providing insights into the efficacy of different model architectures and data balancing methods.

In the semiconductor manufacturing industry, Automatic Defect Classification (ADC) plays an important role in maintaining high wafer inspection quality and reducing yield loss. ADC performance has benefitted from using machine learning (ML) algorithms; however, performance is negatively affected by the data imbalance and limited amounts of training data. Synthetic Minority Oversampling Technique (SMOTE) is an oversampling technique to adjust the skewed class distribution of a dataset so that the bias of the majority class is reduced. This paper shows that applying SMOTE achieved higher accuracy and purity on two imbalanced datasets, consisting of scanning electron microscopy (SEM) images collected with ASML-HMI eP™ and eScan® series inspection tools. The ML models are also less sensitive to the selection of hyperparameters when SMOTE is applied. We also show that better classification results can be obtained with less training samples with SMOTE; we conducted an experiment where a ML model trained on only 25% of samples with SMOTE achieved a higher ADC accuracy and purity performance compared to the same ML model trained on all samples but without SMOTE. In another experiment using a highly imbalanced SEM dataset with very few counts of the defect-of- interest (DOI), the combination of SMOTE and random undersampling of the majority class improves the accuracy by up to 5x while maintaining the same level of purity.

This study investigates the use of machine learning (ML) models combined with a Synthetic Minority Over-sampling Technique (SMOTE) and its variants to predict perioperative pressure injuries (PIs) in an imbalanced dataset. PIs are a significant healthcare problem, often leading to prolonged hospitalization and increased medical costs. Conventional risk assessment scales are limited in their ability to predict PIs accurately, prompting the exploration of ML techniques to address this challenge.We utilized data from 7,292 patients admitted to a tertiary care hospital in Shanghai between May 2017 and July 2023, with a final dataset of 2,972 patients, including 158 with PIs. Seven ML algorithms-Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Extra Trees (ET), K-Nearest Neighbors (KNN), and Decision Trees (DT)-were used in conjunction with SMOTE, SMOTE+ENN, Borderline-SMOTE, ADASYN, and GAN to balance the dataset and improve model performance.Results revealed significant improvements in model performance when SMOTE and its variants were used. For instance, the XGBoost model hadan AUC of 0.996 with SMOTE, compared to 0.800 on raw data. SMOTE+ENN and Borderline-SMOTE further enhanced the models' ability to identify minority classes. External validation indicatedthat XGBoost, RF, and ET exhibited the highest stability and accuracy, with XGBoost having an AUC of 0.977. SHAP analysis revealed that factors such as anesthesia grade, age, and serum albumin levels significantly influenced model predictions.In conclusion, integrating SMOTE with ML algorithms effectively addressed a data imbalance and improved the prediction of perioperative PIs. Future work should focus on refining SMOTE techniques and exploring their application to larger, multi-center datasets to enhance the generalizability of these findings, and especially for diseaseswith a lowincidence.

Employee churn is a significant challenge for organizations, leading to substantial costs associated with recruiting, onboarding, and training new employees. High turnover rates can negatively impact overall productivity, employee morale, and organizational stability. Therefore, accurately predicting employee churn is crucial for companies to implement targeted retention strategies, minimize turnover, and reduce associated expenses. In this study, we leveraged machine learning techniques to predict employee churn using the "HR Analytics" dataset from Kaggle. One of the key challenges in churn prediction is class imbalance, where the number of employees who leave is significantly lower than those who stay. To address this, we applied two data-balancing techniques: Synthetic Minority Over-sampling Technique (SMOTE) and Random Over-Sampling (ROS). We then trained and evaluated four machine learning models Logistic Regression, Random Forest, Decision Tree, and Extreme Gradient Boosting (XGBoost) on the balanced datasets. The F1 scores for the SMOTE-balanced data were: Logistic Regression (0.5990), Random Forest (0.9753), Decision Tree (0.9319), and XGBoost (0.9634). Meanwhile, the ROS-balanced data produced F1 scores of: Logistic Regression (0.5978), Random Forest (0.9760), Decision Tree (0.9475), and XGBoost (0.9703). The results demonstrated that ROS yielded superior performance, particularly for the Random Forest and XGBoost models, leading us to select ROS for further hyperparameter tuning. Using RandomizedSearchCV for optimization, the Random Forest model achieved the highest F1 score of 0.9779. Finally, we deployed the optimized Random Forest model via a Flask API, enabling HR professionals to access a user-friendly web interface for realtime churn prediction. This research highlights the effectiveness of machine learning in HR analytics and underscores the practical benefits of predictive modeling in workforce management, helping organizations proactively address employee retetion challenges.

Drug-Protein Interaction (DPI) identification is crucial in drug discovery. The high dimensionality of drug and protein features poses challenges for accurate interaction prediction, necessitating the use of computational techniques. Docking-based methods rely on 3D structures, while ligand-based methods have limitations such as reliance on known ligands and neglecting protein structure. Therefore, the preferred approach is the chemogenomics-based approach using machine learning, which considers both drug and protein characteristics for DPI prediction. In machine learning, feature selection plays a vital role in improving model performance, reducing overfitting, enhancing interpretability, and making the learning process more efficient. It helps extract meaningful patterns from drug and protein data while eliminating irrelevant or redundant information, resulting in more effective machine-learning models. On the other hand, classification is of great importance as it enables pattern recognition, decision-making, predictive modeling, anomaly detection, data exploration, and automation. It empowers machines to make accurate predictions and facilitates efficient decision-making in DPI prediction. For this research work, protein data was sourced from the KEGG database, while drug data was obtained from the DrugBank data machine-learning base. To address the issue of imbalanced Drug Protein Pairs (DPP), different balancing techniques like Random Over Sampling (ROS), Synthetic Minority Over-sampling Technique (SMOTE), and Adaptive SMOTE were employed. Given the large number of features associated with drugs and proteins, feature selection becomes necessary. Various feature selection methods were evaluated: Correlation, Information Gain (IG), Chi-Square (CS), and Relief. Multiple classification methods, including Support Vector Machines (SVM), Random Forest (RF), Adaboost, and Logistic Regression (LR), were used to predict DPI. Finally, this research identifies the best balancing, feature selection, and classification methods for accurate DPI prediction. This comprehensive approach aims to overcome the limitations of existing methods and provide more reliable and efficient predictions in drug-protein interaction studies.

Machine learning classifiers are increasingly used to create predictive models for pathological complete response (pCR) in breast cancer after neoadjuvant therapy (NAT). Few studies have compared the effectiveness of different ML classifiers. This study evaluated radiomics models based on pre- and post-contrast first-phase T1 weighted images (T1WI) in predicting breast cancer pCR after NAT and compared the performance of ML classifiers. This retrospective study enrolled 281 patients undergoing NAT from the Duke-Breast-Cancer-MRI dataset. Radiomic features were extracted from pre- and post-contrast first-phase T1WI images. The Synthetic Minority Oversampling Technique (SMOTE) was applied, then the dataset was randomly divided into training and validation groups (7:3). The radiomics model was built using selected optimal features. Support vector machine (SVM), random forest (RF), decision tree (DT), k-nearest neighbor (KNN), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM) were classifiers. Receiver operating characteristic curves were used to assess predictive performance. LightGBM performed best in predicting pCR [area under the curve (AUC): 0.823, 95% confidence interval (CI) [0.743-0.902], accuracy 74.0%, sensitivity 85.0%, specificity 67.2%]. During subgroup analysis, RF was most effective in pCR prediction in luminal breast cancers (AUC: 0.914, 95% CI [0.847-0.981], accuracy 87.0%, sensitivity 85.2%, specificity 88.1%). In triple-negative breast cancers, LightGBM performed best (AUC: 0.836, 95% CI [0.708-0.965], accuracy 78.6%, sensitivity 68.2%, specificity 90.0%). The LightGBM-based radiomics model performed best in predicting pCR in patients with breast cancer. RF and LightGBM showed promising results for luminal and triple-negative breast cancers, respectively.

Cervical cancer remains a significant global health burden for women, with approximately 660,000 new cases and 350,000 associated deaths recorded worldwide in 2022. Machine learning methods have shown great promise in advancing timely detection and accurate diagnosis. This investigation compares two widely used oversampling strategies, Synthetic Minority Oversampling Technique (SMOTE) and Adaptive Synthetic Sampling (ADASYN), applied to cervical cancer identification via the XGBoost classifier, paired with Multiple Imputation by Chained Equations (MICE) to handle incomplete data. The dataset consists of cervical cancer risk factors with four diagnostic outcomes: Hinselmann, Schiller, Cytology, and Biopsy, which are treated as independent binary classification tasks rather than a single multilabel classification problem. The process began by preparing a dataset of cervical cancer risk factors through MICE imputation, then applying SMOTE and ADASYN to address class imbalance. The XGBoost model is optimized using Random Search hyperparameter tuning and evaluated across train-test split ratios (50:50, 60:40, 70:30, 80:20, and 90:10) using accuracy, precision (macro, micro, weighted), recall (macro, micro, weighted), F1-score (macro, micro, weighted), and AUC metrics. The results indicated that the XGBoost setup with MICE and SMOTE outperformed the others, achieving 97.1% accuracy, 97.1% mic-precision, 97.1% mic-recall, 97.1% mic-F1, and 97.1% AUC. Meanwhile, the ADASYN-integrated model showed marginally lower results, with 95.4% accuracy, 95.4% micro-precision, 95.4% micro-recall, 95.4% micro-F1, and 55.5% AUC. SMOTE proved more adept at creating evenly distributed synthetic data for the underrepresented group. Overall, this work underscores the value of integrating MICE imputation, SMOTE oversampling, and tuned XGBoost as a reliable approach for cervical cancer detection. These insights pave the way for automated screening tools that can bolster clinical judgment and improve early diagnosis outcomes.

Heart disease is a serious worldwide health issue with wide-reaching effects. Since heart disease is one of the leading causes of mortality worldwide, early detection is crucial. Emerging technologies like Machine Learning (ML) are currently being actively used by the biomedical, healthcare, and health prediction industries. PaRSEL, a new stacking model is proposed in this research, that combines four classifiers, Passive Aggressive Classifier (PAC), Ridge Classifier (RC), Stochastic Gradient Descent Classifier (SGDC), and eXtreme Gradient Boosting (XGBoost), at the base layer, and LogitBoost is deployed for the final predictions at the meta layer. The imbalanced and irrelevant features in the data increase the complexity of the classification models. The dimensionality reduction and data balancing approaches are considered very important for lowering costs and increasing the accuracy of the model. In PaRSEL, three dimensionality reduction techniques, Recursive Feature Elimination (RFE), Linear Discriminant Analysis (LDA), and Factor Analysis (FA), are used to reduce the dimensionality and select the most relevant features for the diagnosis of heart disease. Furthermore, eight balancing techniques, Proximity Weighted Random Affine Shadowsampling (ProWRAS), Localized Randomized Affine Shadowsampling (LoRAS), Random Over Sampling (ROS), Adaptive Synthetic (ADASYN), Synthetic Minority Oversampling Technique (SMOTE), Borderline SMOTE (B-SMOTE), Majority Weighted Minority Oversampling Technique (MWMOTE) and Random Walk Oversampling (RWOS), are used to deal with the imbalanced nature of the dataset. The performance of PaRSEL is compared with the other standalone classifiers using different performance measures like accuracy, F1-score, precision, recall and AUC-ROC score. Our proposed model achieves 97% accuracy, 80% F1-score, precision is greater than 90%, 67% recall, and 98% AUC-ROC score. This shows that PaRSEL outperforms other standalone classifiers in terms of heart disease prediction. Additionally, we deploy SHapley Additive exPlanations (SHAP) on our proposed model. It helps to understand the internal working of the model. It illustrates how much influence a classifier has on the final prediction outcome.

Depression and anxiety are significant global public health, their comorbidity share similar clinical symptoms with heterogeneous progression patterns for patients. The early and accurate prediction of anxiety, depression, and comorbidity of anxiety and depression is essential for timely discrimination of high-risk patients, individual treatment strategies, and hierarchical management. This study aimed to develop and validate an interpretable, multi-task prediction model using heart rate variability (HRV) for discriminating anxiety alone, depression alone, and comorbid patients with psychiatric disorders. Patients diagnosed with anxiety or depression by trained psychiatrists were retrospectively enrolled from the Department of Psychiatry, the Affiliated Hospital of Southwest Medical University, between March 2024 and April 2025. The Synthetic Minority Over-sampling Technique (SMOTE) was applied to address class imbalance, and the Recursive feature elimination (RFE) was used for feature selection. Six machine-learning algorithms, including logistic regression (LR), support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), k-nearest neighbor (KNN), and naïve Bayes (NB), were adopt to develop multi-task prediction models for discriminating psychiatric comorbidities. Model performances were evaluated by receiver operating characteristic (ROC) curves, accuracy, sensitivity, specificity, F1 score, and Matthews correlation coefficient (MCC). The SHapley Additive exPlanation (SHAP) method was used to visualize the feature contributions and interpret the best model. A total of 546 patients, 114 with anxiety group(20.88%), 100 with depression group (18.32%), and 332 with comorbidity of anxiety and depression (60.81%), were included in this study. The ML-based models were developed for each task using the eleven contributing features. Among the six ML models, the XGBoost achieved the best performance across all tasks. In the validation set, this model achieved the AUCs of 0.8341 (95% confidence interval [CI]:0.761–0.907) for the task of anxiety versus comorbidity of anxiety and depression, and 0.8458(95% CI:0.775–0.917) for the task of depression versus comorbidity of anxiety and depression, respectively. The clinical utility was evaluated using Decision curve analysis (DCA) and calibration curves. The proposed model has potential clinical application prospects and may assist clinicians to improve the accuracy and efficiency of diagnosis, and thereby facilitating timely appropriate intervention measures for patients with psychiatric disorders. Not applicable.

<p>Problem in imbalanced data is very common in classification or identification. The problem is raised when the number of instances of one class far exceeds the other. In the previous research, our DNA barcode-based Identification System of Tuna and Mackerel was developed in imbalanced dataset. The number of samples of Tuna and Mackerel were much more than those of other fish samples. Therefore, the accuracy of the classification model was probably still in bias. This research aimed at employing Synthetic Minority Oversampling Technique (SMOTE) to yield balanced dataset. We used kmers frequencies from DNA barcode sequences as features and Support Vector Machine (SVM) as classification method. In this research we used trinucleotide (3-mers) and tetranucleotide (4-mers). The training dataset was taken from Barcode of Life Database (BOLD). For evaluating the model, we compared the accuracy of model using SMOTE and without SMOTE in order to classify DNA barcode sequences which is taken from Department of Aquatic Product Technology, Bogor Agricultural University. The results showed that the accuracy of the model in the species level using SMOTE was 7% and 13% higher than those of non-SMOTE for trinucleotide (3-mers) and tetranucleotide (4-mers), respectively. It is expected that the use of SMOTE, as one of data balancing technique, could increase the accuracy of DNA barcode based fish classification system, particularly in the species level which is difficult to be identified.</p>

Parkinson’s disease (PD) is a progressive neurological disorder that affects millions globally, posing significant challenges in early and accurate diagnosis. Recent advancements in machine learning (ML) offer promising approaches for addressing these challenges by enabling more precise and efficient PD predictions. This paper proposes an enhanced ML framework for PD prediction, integrating data balancing, feature selection, and explainable AI techniques. We evaluate nine different ML algorithms using a dataset of clinical and voice features. To address the class imbalance, we employ the Synthetic Minority Oversampling Technique (SMOTE) and NearMiss, comparing results to an imbalanced baseline. Feature engineering approaches, including Featurewiz, Tree based Feature Importance and the chi-square test, are utilized to identify key predictive features such as Pitch Period Entropy (PPE), Noise-to-Harmonic Ratio (NHR), and other voice biomarkers. Explainable AI (XAI) techniques (SHAP and LIME) interpret model decision-making and highlight influential features. The best-performing model, KNN with SMOTE, achieved 92% accuracy, F1-score 0.94, and a G-Mean of 0.95—demonstrating balanced, reliable PD detection. While some models achieved higher accuracy on imbalanced data (up to 97%), their performance lacked sensitivity and balance. Our findings suggest that combining SMOTE with feature engineering and XAI substantially enhances model fairness, performance, and interpretability. This research advances PD prediction by providing an accurate and interpretable ML-based diagnostic tool to support early diagnosis and better patient management.

Stroke is one of the leading causes of death worldwide, creating an urgent need for effective early detection systems, particularly because conventional methods often struggle with class imbalance and produce biased evaluations. Previous studies have primarily focused on accuracy while overlooking model consistency, data pre-processing quality, and probability-based evaluation. This study evaluates model performance under three conditions: original data using extreme gradient boosting (XGBoost) with scale_pos_weight, original data using the easy ensemble classifier, and class-balanced data generated using random oversampling (ROS), adaptive synthetic sampling (ADASYN), and synthetic minority over-sampling technique (SMOTE). Each model underwent missing value handling, normalization, feature preparation, and hyperparameter optimization using grid search. Performance was assessed using area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), confidence intervals, calibration curves, Shapley additive explanations (SHAP), decision curve analysis (DCA), and external validation. The results demonstrate that data resampling significantly improves performance, with the XGBoost-SMOTE combination achieving the best results, including an accuracy of 0.99, AUROC of 0.998, and AUPRC of 0.986, outperforming the other approaches. This method provides more consistent and balanced predictions, supporting the application of artificial intelligence for early stroke risk identification.

Prediction of the stage of cancer plays an important role in planning the course of treatment and has been largely reliant on imaging tools which do not capture molecular events that cause cancer progression. Gene-expression data-based analyses are able to identify these events, allowing RNA-sequence and microarray cancer data to be used for cancer analyses. Breast cancer is the most common cancer worldwide, and is classified into four stages - stages 1, 2, 3, and 4 [2]. While machine learning models have previously been explored to perform stage classification with limited success, multi-class stage classification has not had significant progress. There is a need for improved multi-class classification models, such as by investigating deep learning models. Gene-expression-based cancer data is characterised by the small size of available datasets, class imbalance, and high dimensionality. Class balancing methods must be applied to the dataset. Since all the genes are not necessary for stage prediction, retaining only the necessary genes can improve classification accuracy. The breast cancer samples are to be classified into 4 classes of stages 1 to 4. Invasive ductal carcinoma breast cancer samples are obtained from The Cancer Genome Atlas (TCGA) and Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) datasets and combined. Two class balancing techniques are explored, synthetic minority oversampling technique (SMOTE) and SMOTE followed by random undersampling. A hybrid feature selection pipeline is proposed, with three pipelines explored involving combinations of filter and embedded feature selection methods: Pipeline 1 - minimum-redundancy maximum-relevancy (mRMR) and correlation feature selection (CFS), Pipeline 2 - mRMR, mutual information (MI) and CFS, and Pipeline 3 - mRMR and support vector machine-recursive feature elimination (SVM-RFE). The classification is done using deep learning models, namely deep neural network, convolutional neural network, recurrent neural network, a modified deep neural network, and an AutoKeras generated model. Classification performance post class-balancing and various feature selection techniques show marked improvement over classification prior to feature selection. The best multiclass classification was found to be by a deep neural network post SMOTE and random undersampling, and feature selection using mRMR and recursive feature elimination, with a Cohen-Kappa score of 0.303 and a classification accuracy of 53.1%. For binary classification into early and late-stage cancer, the best performance is obtained by a modified deep neural network (DNN) post SMOTE and random undersampling, and feature selection using mRMR and recursive feature elimination, with an accuracy of 81.0% and a Cohen-Kappa score (CKS) of 0.280. This pipeline also showed improved multiclass classification performance on neuroblastoma cancer data, with a best area under the receiver operating characteristic (auROC) curve score of 0.872, as compared to 0.71 obtained in previous work, an improvement of 22.81%. The results and analysis reveal that feature selection techniques play a vital role in gene-expression data-based classification, and the proposed hybrid feature selection pipeline improves classification performance. Multi-class classification is possible using deep learning models, though further improvement particularly in late-stage classification is necessary and should be explored further.