Diabetes Risk Modeling through Tabular-to-Image Transformations and Ensemble Learning

Hybrid Sampling and Ensemble Learning for Food Safety Sampling Inspection Classification.

Traffic safety is a critical issue in urban transportation systems, with crash severity predictions playing a key role in mitigating road traffic injuries and fatalities. This study addresses the research question: how can transforming tabular crash data into image-based representations using deep learning techniques improve the prediction of crash severity, while effectively managing class imbalances? For analysis, 63,745 barrier crashes that occurred in Texas from 2017–2022 were examined. Using a sample database of barrier crash incidents, we introduce a novel approach that leverages Convolutional Neural Networks (CNNs) to convert traditional tabular crash data into image representations by employing the DeepInsight technique, which maps feature similarity into structured two-dimensional grids, enhancing the accuracy and effectiveness of prediction models. Generative Adversarial Networks (GANs) were used for data augmentation, addressing issues of class imbalances with innovative sampling techniques including under-sampling, over-sampling, and balanced sampling. Key findings indicate that balanced datasets significantly improve predictive performance, enhancing generalizability across diverse crash scenarios. Moreover, under- and over-sampling strategies offer a refined trade-off among precision, recall, and F1-score, while the over-sampling method distinctly excels in practical model performance by reducing misclassification rates and improving overall accuracy. Comparative evaluation against an XGBoost baseline on identical balanced data highlights the ability of the image-based deep learning framework to capture complex crash patterns. These findings suggest that integrating feature-similarity–based tabular-to-image transformation with deep learning and balanced data sampling can significantly improve crash severity classification.

Regularization is an important technique for developing deep learning models to improve generalization and reduce overfitting. This study evaluated the effect of regularization on the performance of neural network models in regression prediction tasks using earthquake data. We compare Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), and Feedforward Neural Network (FNN) architectures with L2 and Dropout regularization. The experimental results show that MLP without regularization achieved the best performance (RMSE: 0.500, MAE: 0.380, R²: 0.625), although prone to overfitting. CNN performed poorly on tabular data, while FNN showed marginal improvement with deeper layers. The novelty of this study lies in a comparative evaluation of regularization strategies across multiple architectures for earthquake regression prediction, highlighting practical implications for early warning systems.

Detection of breast cancer remains an important global health issue, particularly in patients with dense breast tissue. Despite wide use of machine learning for breast cancer diagnosis, many existing studies rely on single-model evaluations and lack rigorous statistical validation, while transformer-based models remain insufficiently explored for structured tabular biomedical data. Using the Wisconsin Breast Cancer Diagnostic (WBCD) dataset, this study presents a comparative analysis of machine learning, deep learning, and transformer-based models for binary classification of breast masses as benign or malignant. A robust preprocessing and feature engineering pipeline was applied, and Support Vector Machine (SVM), Random Forest (RF), Multi-Layer Perceptron (MLP), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and BERT-based models were evaluated using a stratified 66–34% train–test split and stratified 5-fold cross-validation. Performance was assessed using standard classification metrics, Jaccard Index, AUC-ROC, false negative rate, and statistical significance testing (paired t-tests and Bayesian analysis). Three-layer RNN and LSTM models achieved the highest accuracy of 98.25%, showing statistically significant improvement over SVM (p = 0.042), while the optimized SVM demonstrated strong performance with superior computational efficiency. BERT-based models yielded lower accuracy (92.98%), reflecting domain mismatch between language-pretrained transformers and numerical tabular data. These findings indicate that deeper recurrent architectures provide superior diagnostic performance, while SVM remains a practical and efficient alternative for clinical deployment.

Climate change poses significant challenges to agriculture, leading to increased crop damage owing to extreme weather conditions. Detecting and analyzing such damage is crucial for mitigating its effects on crop yield. This study proposes a novel autoencoder (AE)-based model, termed “Memory Ganomaly,” designed to detect and analyze weather-induced crop damage under conditions of significant class imbalance. The model integrates memory modules into the Ganomaly architecture, thereby enhancing its ability to identify anomalies by focusing on normal (undamaged) states. The proposed model was evaluated using apple and peach datasets, which included both damaged and undamaged images, and was compared with existing robust Convolutional neural network (CNN) models (ResNet-50, EfficientNet-B3, and ResNeXt-50) and AE models (Ganomaly and MemAE). Although these CNN models are not the latest technologies, they are still highly effective for image classification tasks and are deemed suitable for comparative analyses. The results showed that CNN and Transformer baselines achieved very high overall accuracy (94–98%) but completely failed to identify damaged samples, with precision and recall equal to zero under severe class imbalance. Few-shot learning partially alleviated this issue (up to 75.1% recall in the 20-shot setting for the apple dataset) but still lagged behind AE-based approaches in terms of accuracy and precision. In contrast, the proposed Memory Ganomaly delivered a more balanced performance across accuracy, precision, and recall (Apple: 80.32% accuracy, 79.4% precision, 79.1% recall; Peach: 81.06% accuracy, 83.23% precision, 80.3% recall), outperforming AE baselines in precision and recall while maintaining comparable accuracy. This study concludes that the Memory Ganomaly model offers a robust solution for detecting anomalies in agricultural datasets, where data imbalance is prevalent, and suggests its potential for broader applications in agricultural monitoring and beyond. While both Ganomaly and MemAE have shown promise in anomaly detection, they suffer from limitations—Ganomaly often lacks long-term pattern recall, and MemAE may miss contextual cues. Our proposed Memory Ganomaly integrates the strengths of both, leveraging contextual reconstruction with pattern recall to enhance detection of subtle weather-related anomalies under class imbalance.

Backgroundand Purpose:This study aimed to develop and evaluate an efficient method to automatically segment T1- and T2-weighted brain magnetic resonance imaging (MRI) images. We specifically compared the segmentation performance of individual convolutional neural network (CNN) models against an ensemble approach to advance the accuracy of MRI-guided radiotherapy (RT) planning.Materials and Methods. The evaluation was conducted on a private clinical dataset and a publicly available dataset (HaN-Seg). Anonymized MRI data from 55 brain cancer patients, including T1-weighted, T1-weighted with contrast, and T2-weighted images, were used in the clinical dataset. We employed an EDL strategy that integrated five independently trained 2D neural networks, each tailored for precise segmentation of tumors and organs at risk (OARs) in the MRI scans. Class probabilities were obtained by averaging the final layer activations (Softmax outputs) from the five networks using a weighted-average method, which were then converted into discrete labels. Segmentation performance was evaluated using the Dice similarity coefficient (DSC) and Hausdorff distance at 95% (HD95). The EDL model was also tested on the HaN-Seg public dataset for comparison.Results. The EDL model demonstrated superior segmentation performance on both the clinical and public datasets. For the clinical dataset, the ensemble approach achieved an average DSC of 0.7 ± 0.2 and HD95 of 4.5 ± 2.5 mm across all segmentations, significantly outperforming individual networks which yielded DSC values ≤0.6 and HD95 values ≥14 mm. Similar improvements were observed in the HaN-Seg public dataset.Conclusions. Our study shows that the EDL model consistently outperforms individual CNN networks in both clinical and public datasets, demonstrating the potential of ensemble learning to enhance segmentation accuracy. These findings underscore the value of the EDL approach for clinical applications, particularly in MRI-guided RT planning.

BackgroundAccurate and interpretable models are essential for clinical decision-making, where predictions can directly impact patient care. Machine learning (ML) survival methods can handle complex multidimensional data and achieve high accuracy but require post-hoc explanations. Traditional models such as the Cox Proportional Hazards Model (Cox-PH) are less flexible, but fast, stable, and intrinsically transparent. Moreover, ML does not always outperform Cox-PH in clinical settings, warranting a diligent model validation. We aimed to develop a set of R functions to help explore the limits of Cox-PH compared to the tree-based and deep learning survival models for clinical prediction modelling, employing ensemble learning and nested cross-validation. MethodsWe developed a set of R functions, publicly available as the package “survcompare”. It supports Cox-PH and Cox-Lasso, and Survival Random Forest (SRF) and DeepHit are the ML alternatives, along with the ensemble methods integrating Cox-PH with SRF or DeepHit designed to isolate the marginal value of ML. The package performs a repeated nested cross-validation and tests for statistical significance of the ML’s superiority using the survival-specific performance metrics, the concordance index, time-dependent AUC-ROC and calibration slope.To get practical insights, we applied this methodology to clinical and simulated datasets with varying complexities and sizes. ResultsIn simulated data with non-linearities or interactions, ML models outperformed Cox-PH at sample sizes ≥ 500. ML superiority was also observed in imaging and high-dimensional clinical data. However, for tabular clinical data, the performance gains of ML were minimal; in some cases, regularised Cox-Lasso recovered much of the ML’s performance advantage with significantly faster computations. Ensemble methods combining Cox-PH and ML predictions were instrumental in quantifying Cox-PH’s limits and improving ML calibration. Traditional models like Cox-PH or Cox-Lasso should not be overlooked while developing clinical predictive models from tabular data or data of limited size. ConclusionOur package offers researchers a framework and practical tool for evaluating the accuracy-interpretability trade-off, helping make informed decisions about model selection.

Accurate prediction of diabetes based on clinical and demographic indicators is essential, as early prediction of this chronic metabolic disorder plays a critical role in preventing long-term organ complications. However, existing research continues to face significant challenges, including pronounced class imbalance, the scarcity of large and diverse datasets, and limited integration of explainable artificial intelligence. This research compares several ensemble learning methods (Decision Tree, Random Forest, AdaBoost, Gradient Boosting, Histogram based Gradient Boosting, extremely randomized Trees, and XGBoost) on a large imbalanced dataset (87,664 negative vs. 8,482 positive samples). To mitigate imbalance, we evaluate six resampling approaches including (Random Over Sampling, Random Under Sampling, Synthetic Minority Over-sampling, Adaptive Synthetic Sampling, Tomek Links Removal Sampling and (Synthetic Minority Over-sampling with Edited Nearest Neighbors). We assess models using metrics robust to class imbalance (precision, recall, F1, AUC-ROC, and AUC-PR) and calibration measures. The Extra Trees classifier achieved the highest measured accuracy (0.994); with Random Over Sampling for balancing dataset. also, these results were compared with several previous works and number of machine learning algorithms, and the results showed superiority. Explainability is performed at both global and local levels: permutation and SHAP for global feature importance, and (Local Interpretable Model-Agnostic Explanations) force plots for instance-level reasoning. however, we analyzed this result using sensitivity, specificity, PR-AUC and calibration, we report detailed experiments showing how resampling method, hyperparameter tuning, and stratified validation influence performance. Finally, we provide clinical-relevant insights from SHAP analyses and discuss limitations and future directions for deploying interpretable models in screening workflows.

Tabular data within enterprises or open data repositories provide a huge opportunity for feature augmentation. Using these data sources to augment training data often boosts model performance, which is crucial in data-centric AutoML systems. Recent works on automatic feature augmentation have limited capabilities in utilizing useful features that cannot be joined with the base table without connecting through intermediate tables. We present Featpilot, a novel framework that explores and integrates high-quality features in tabular data for ML models. Featpilot evaluates a candidate feature from two aspects: (1) the efficacy of a join path connecting the feature to the base table and (2) the intrinsic value of a feature towards an ML task. Featpilot efficiently identifies high-quality features and their optimized join paths to augment the base table. Our experimental results show that Featpilot achieves up to a 10.27% improvement in ML model performance compared to state-of-the-art solutions across six public datasets.

Objectives: To propose a method for imputing missing values within non-linear datasets with class imbalance issues. Methods: The proposed methodology includes two stages of imputation through the Polynomial with Backfill (PoBa) process. This imputation technique applies quadratic polynomial interpolation for individual class type missing value estimation as well as implementing a backfill strategy to complete remaining gaps. The evaluation process utilized six ML models to assess dataset imputations under different non-linear and class imbalance conditions within two publicly available datasets. Findings: The PoBa method showed superior performance than HPCNN baseline in all the metric evaluations for different degrees of missingness. The Wine Quality dataset received an accuracy score of 0.807 and MAE of 0.193 and RMSE of 0.439 when feeding PoBa to Multi-Layer Perceptron (MLP) under 30% missing data conditions. The PoBa method reached 0.950 accuracy along with 0.050 MAE and 0.224 RMSE in the heart disease dataset despite its imbalanced characteristics. The experiments demonstrate that PoBa effectively preserves data organization and improves predictive performance while handling large real-world datasets. Novelty: The method proposes a class-specific combination of polynomial interpolation with backfill to handle missing data in an efficient and interpretable way. Keywords: Missing Data, Polynomial Interpolation, Backfill Imputation, Machine Learning, Class Imbalance

Soft reordering one-dimensional convolutional neural network for credit scoring

Objective: Alzheimer's disease (AD) is a neurodegenerative disorder that severely impairs cognitive function across various age groups, ranging from early to late sixties. It progresses from mild to severe stages, so an accurate diagnostic tool is necessary for effective intervention and treatment planning. Methods: This work proposes a novel AD classification architecture that integrates depthwise separable convolutional layers with traditional convolutional layers to efficiently extract features from structural magnetic resonance imaging (sMRI) scans. This model benefits from excellent feature extraction and lightweight operation, which reduces the number of parameters without compromising accuracy. The model learns from scratch with optimized weight initialization, resulting in faster convergence and improved generalization. However, medical imaging datasets contain class imbalance as a major challenge, which often results in biased models with poor generalization to the underrepresented disease stages. A hybrid sampling approach combining SMOTE (synthetic minority oversampling technique) with the ENN (edited nearest neighbors) effectively handles the complications of class imbalance issue inherent in the datasets. An explainable activation space occlusion sensitivity map (ASOP) pixel attribution method is employed to highlight the critical regions of input images that influence the classification decisions across different stages of AD. Results and Conclusions: The proposed model outperformed several state-of-the-art transfer learning architectures, including VGG19, DenseNet201, EfficientNetV2S, MobileNet, ResNet152, InceptionV3, and Xception. It achieves noteworthy results in disease stage classification, with an accuracy of 98.87%, an F1 score of 98.86%, a precision of 98.80%, and recall of 98.69%. These results demonstrate the effectiveness of the proposed model for classifying stages of AD progression.

Graph convolutional network – Long short term memory neural network- multi layer perceptron- Gaussian progress regression model: A new deep learning model for predicting ozone concertation

Emotion recognition plays a pivotal role in advancing Human-Computer Interaction (HCI) by enabling systems to understand and respond to human emotional states. Among various modalities, Speech Emotion Recognition (SER) stands out as a non-invasive, cost-effective, and temporally efficient approach for detecting emotions. This study leverages the RAVDESS dataset, which provides high-quality audio recordings for emotion classification. The proposed methodology involves preprocessing audio signals to remove noise, extracting temporal and spectral features in both time and frequency domains, and implementing machine learning models for multi-class emotion classification. The study evaluates and compares the performance of several classification models, including Random Forest (RF), Multilayer Perceptron (MLP), Support Vector Machine (SVM), Convolutional Neural Networks (CNN), and Decision Trees (DT). Experimental results demonstrate promising accuracy levels, highlighting the potential of these machine learning techniques in SER. This research contributes to the integration of SER within Brain-Computer Interfaces (BCI) and other emotion-aware applications, paving the way for enhanced interactive systems. Keywords – Speech Emotion Recognition -Machine Learning - Random Forest -Multilayer Perceptron - Support Vector Machine - Convolutional Neural Networks - Decision Tree - RAVDESS Dataset

Software Fault Prediction (SFP) is the process of predicting fault-prone software constructs during the initial phases of software development. Deep Neural Networks (DNN) have been hugely successful in the field of computer vision, audio, etc., where the input data is correlated spatially and temporally. In contrast, SFP operates on tabular data, rows of software metrics that lack the inherent spatial structure exploited by convolutional architectures in image domains. The authors have tried to remodel the most successful Convolutional Neural Network (CNN) for tabular data. A novel framework is proposed employing a tree-based feature selection technique, LSelect, to find the most significant features and an adaptive 1-dimensional Convolutional Neural Network (ACNN) for the classification task, which selects an optimal learning rate automatically. ACNN converts the tabular data (1-D) into 2-D using adaptive pooling layers, thereby forming an image from 1-D data. The framework classification results (Area under Curve) are compared with nine state-of-the-art algorithms, such as XGBoost, LightGBM, etc., and performance is validated using the Bayesian Signed Rank Test. It is found that the proposed framework performs comparably with the state-of-the-art methods with reduced model complexity. Also, the LSelect feature selection technique improves average model performance by 1.3%.