Robust Classification Research Articles

FusionNet remote a hybrid deep learning ensemble model for remote image classification in multispectral images

Remote sensing plays a vital role in various industries, including smart cities, agriculture, and environmental monitoring, by capturing multispectral images of the Earth’s surface and analyzing its features. Accurate classification of these images is essential for extracting valuable information and making informed decisions. However, traditional image classification models, such as Convolutional Neural Networks (CNNs), often struggle to fully capture the complex spectral and spatial information inherent in multispectral images. Addressing this research gap, we propose FusionNet-Remote, a novel hybrid deep learning ensemble model that integrates CNNs with Random Forests (RF) to enhance the accuracy and robustness of remote image classification. The primary objective of this study is to develop a model that combines the strengths of CNNs for spatial feature extraction with the robust classification capabilities of RFs. Through comprehensive evaluations using LANDSAT data, FusionNet-Remote demonstrates superior performance compared to existing models, achieving an outstanding training accuracy of 99.8% and a testing accuracy of 98.7%. The model also exhibits the lowest training loss (0.0001) and testing loss (0.0043), significantly outperforming traditional models such as standalone CNNs, VGG16, and Inceptionv3. These results highlight the effectiveness of the hybrid ensemble approach in overcoming the limitations of conventional methods. This research contributes to the field by introducing an innovative ensemble technique that significantly improves remote image classification accuracy and reliability. The model’s high performance underscores its potential for critical applications such as land cover mapping, vegetation analysis, and disaster monitoring. Future work will explore the integration of more advanced CNN architectures and alternative fusion techniques to further enhance the model's performance and adaptability across different remote sensing scenarios.

Ensemble-based customer churn prediction in banking: a voting classifier approach for improved client retention using demographic and behavioral data

Customer turnover is a crucial issue in banking since maintained profitability depends on keeping clients. This work aims to categorize consumer turnover in banks by using a new ensemble approach combining many machine learning methods, hence enhancing churn prediction models. Using a comprehensive dataset including demographic, financial, and behavioral data—such as credit score, account balance, tenure, and activity levels—the study employs the goal variable revealing if a customer has left the bank. The study starts with univariate, bivariate, and multivariate feature exploration and subsequently uses the Interquartile Range (IQR) approach to identify outliers thereby improving the data quality. Five models—K-Nearest Neighbors, Support Vector Classifier, Random Forest, Decision Tree, and XGBoost—a Voting Classifier ensemble—are used to estimate project churn. Building upon all the strengths of each model, this approach improves the prediction of classification and provides a balanced and highly robust classification system. The applied approaches are K-Nearest Neighbors (KNN), Support Vector Classifier (SVC), Random Forest, Decision Tree, and XGBoost within a Voting Classifier configuration. The performance of the Voting Classifier without SMOTE yields the following results: Accuracy: 0.87, precision: 0.87, recall: 0.80, and F1-Score: 0.87. The proposed model that extend the base model using SMOTE (Synthetic Minority Over-sampling Technique), yields a higher prediction accuracy of 0.90, precision of 0.90, recall of 0.90 and F1-Score of 0.90. This enhancement is proving the efficiency of SMOTE to handle the class imbalance problem in order to render the churn prediction more balanced and reliable system. The proposed approach assures a reliable solution to the strategies to retain the customers in the banking organisations.

Intelligent Symptom Analysis and Report Interpretation for Thyroid Disorder Prediction Using Machine Learning

The thyroid gland, situated at the front of the neck, is essential for producing hormones that regulate many of the body’s critical functions. This paper presents a machine learning-based system aimed at enhancing early detection and providing personalized care for thyroid disorders. Leveraging the Random Forest algorithm chosen for its robust classification abilities the system incorporates two key models: an initial assessment tool that examines self-reported symptoms to determine if further diagnostic testing is recommended, and a diagnostic model that interprets test results to identify specific thyroid disorders. Additionally, the system provides customized diet and exercise suggestions. With an accuracy of 85%, the system effectively supports early symptom analysis and accurate diagnosis. By combining symptom evaluation with tailored health guidance, this approach enables timely detection, accurate diagnostics, and effective management of thyroid conditions, offering personalized insights for patients and support for healthcare providers. Key Words: Machine Learning, Random Forest, Thyroid Disease , Data Preprocessing, Healthcare informatics

Fake News Detection and Classification: A Comparative Study of Convolutional Neural Networks, Large Language Models, and Natural Language Processing Models

In an era where fake news detection has become a pressing issue due to its profound impacts on public opinion, democracy, and social trust, accurately identifying and classifying false information is a critical challenge. In this study, the effectiveness is investigated of advanced machine learning models—convolutional neural networks (CNNs), bidirectional encoder representations from transformers (BERT), and generative pre-trained transformers (GPTs)—for robust fake news classification. Each model brings unique strengths to the task, from CNNs’ pattern recognition capabilities to BERT and GPTs’ contextual understanding in the embedding space. Our results demonstrate that the fine-tuned GPT-4 Omni models achieve 98.6% accuracy, significantly outperforming traditional models like CNNs, which achieved only 58.6%. Notably, the smaller GPT-4o mini model performed comparably to its larger counterpart, highlighting the cost-effectiveness of smaller models for specialized tasks. These findings emphasize the importance of fine-tuning large language models (LLMs) to optimize the performance for complex tasks such as fake news classifier development, where capturing subtle contextual relationships in text is crucial. However, challenges such as computational costs and suboptimal outcomes in zero-shot classification persist, particularly when distinguishing fake content from legitimate information. By highlighting the practical application of fine-tuned LLMs and exploring the potential of few-shot learning for fake news detection, this research provides valuable insights for news organizations seeking to implement scalable and accurate solutions. Ultimately, this work contributes to fostering transparency and integrity in journalism through innovative AI-driven methods for fake news classification and automated fake news classifier systems.

A robust multimodal brain MRI-based diagnostic model for migraine: validation across different migraine phases and longitudinal follow-up data

Inter-individual variability in symptoms and the dynamic nature of brain pathophysiology present significant challenges in constructing a robust diagnostic model for migraine. In this study, we aimed to integrate different types of magnetic resonance imaging (MRI), providing structural and functional information, and develop a robust machine learning model that classifies migraine patients from healthy controls by testing multiple combinations of hyperparameters to ensure stability across different migraine phases and longitudinally repeated data. Specifically, we constructed a diagnostic model to classify patients with episodic migraine from healthy controls, and validated its performance across ictal and interictal phases, as well as in a longitudinal setting. We obtained T1-weighted and resting-state functional MRI data from 50 patients with episodic migraine and 50 age- and sex-matched healthy controls, with follow-up data collected after one year. Morphological features, including cortical thickness, curvature, and sulcal depth, and functional connectivity features, such as low-dimensional representation of functional connectivity (gradient), degree centrality, and betweenness centrality, were utilized. We employed a regularization-based feature selection method combined with a random forest classifier to construct a diagnostic model. By testing the models with varying feature combinations, penalty terms, and spatial granularities within a strict cross-validation framework, we found that the combination of curvature, sulcal depth, cortical thickness, and functional gradient achieved a robust classification performance. The model performance was assessed using the test dataset and achieved 87% accuracy and 0.94 area under the curve (AUC) at distinguishing migraine patients from healthy controls, with 85%, 0.97 and 84%, 0.93 during the interictal and ictal/peri-ictal phases, respectively. When validated using follow-up data, which was not included during model training, the model achieved 91%, 94%, 89% accuracies and 0.96, 0.94, 0.98 AUC for the total, interictal, and ictal/peri-ictal phases, respectively, confirming its robustness. Feature importance and clinical association analyses exhibited that the somatomotor, limbic, and default mode regions could be reliable markers of migraine. Our findings, which demonstrate a robust diagnostic performance using multimodal MRI features and a machine-learning framework, may offer a valuable approach for clinical diagnosis across diverse cohorts and help alleviate the decision-making burden for clinicians.

Classification and analysis of chronic health conditions influencing dementia progression using machine learning methods

Dementia, a progressive neurodegenerative condition, poses significant challenges to global healthcare systems due to its rising prevalence and complexity in diagnosis. This study employs a machine learning-based methodology to categorize dementia-related data and forecast disease progression, enabling early diagnosis and personalized treatment planning. Using a structured dataset comprising 24 features and 2230 entries, data preprocessing was conducted to address missing values, encode categorical variables, and normalize numerical features. Exploratory data analysis revealed key insights, including a strong inverse correlation between cognitive test scores and dementia severity. A Random Forest Classifier was trained to predict dementia presence, achieving an accuracy of 91.40%. Feature importance analysis identified "Cognitive_Test_Scores," "Prescription_Memantine," and "Dosage in mg" as the most critical predictors, aligning with clinical evidence. The confusion matrix highlighted high predictive accuracy but revealed misclassifications, particularly false negatives, which are critical in medical contexts. This study underscores the potential of machine learning in dementia diagnosis and management. By identifying significant predictors and achieving robust classification performance, the research provides a foundation for integrating artificial intelligence into healthcare systems to enhance diagnostic accuracy, optimize treatment plans, and improve patient outcomes. Further refinement and validation using larger datasets are recommended for broader applicability in clinical settings.

AIoT-Driven Leak Detection in Real Water Networks Using Hydrophones

AbstractAcoustic sensing technology is a familiar approach to detect leakage in urban water networks. Critical issues like false alarms, difficult leak locations, missed leaks, unknown site conditions, and high repair costs are still prevalent. The situation warrants developing a more sophisticated and efficient leak detection approach in real water networks. Hydrophone based acoustic technology has a strong promise for high precision detection of leaks. However, AIoT approach using hydroacoustic data for real water leak detection are rarely reported. The current study, therefore, proposes an integrated signal analysis and machine learning-based ensemble model for leak detection using a hydrophone-based smart IoT system. The results show that the most significant features are peak frequency and maximum amplitude. Random forest is the most robust classifier for cost effective long-term monitoring, and the proposed voting ensemble classifies leaks and no leaks with high accuracy on both unseen data and new sites. Specifically, proposed models have very few alarms and missed leaks are reported, a significant problem in models developed using accelerometers and noise loggers. The study shows a significant contribution to the domain of leak detection for real urban water networks.

Developing a Sleep Algxorithm to Support a Digital Medicine System: Noninterventional, Observational Sleep Study.

Sleep-wake patterns are important behavioral biomarkers for patients with serious mental illness (SMI), providing insight into their well-being. The gold standard for monitoring sleep is polysomnography (PSG), which requires a sleep lab facility; however, advances in wearable sensor technology allow for real-world sleep-wake monitoring. The goal of this study was to develop a PSG-validated sleep algorithm using accelerometer (ACC) and electrocardiogram (ECG) data from a wearable patch to accurately quantify sleep in a real-world setting. In this noninterventional, nonsignificant-risk, abbreviated investigational device exemption, single-site study, participants wore the reusable wearable sensor version 2 (RW2) patch. The RW2 patch is part of a digital medicine system (aripiprazole with sensor) designed to provide objective records of medication ingestion for patients with schizophrenia, bipolar I disorder, and major depressive disorder. This study developed a sleep algorithm from patch data and did not contain any study-related or digitized medication. Patch-acquired ACC and ECG data were compared against PSG data to build machine learning classification models to distinguish periods of wake from sleep. The PSG data provided sleep stage classifications at 30-second intervals, which were combined into 5-minute windows and labeled as sleep or wake based on the majority of sleep stages within the window. ACC and ECG features were derived for each 5-minute window. The algorithm that most accurately predicted sleep parameters against PSG data was compared to commercially available wearable devices to further benchmark model performance. Of 80 participants enrolled, 60 had at least 1 night of analyzable ACC and ECG data (25 healthy volunteers and 35 participants with diagnosed SMI). Overall, 10,574 valid 5-minute windows were identified (5854 from participants with SMI), and 84% (n=8830) were classified as greater than half sleep. Of the 3 models tested, the conditional random field algorithm provided the most robust sleep-wake classification. Performance was comparable to the middle 50% of commercial devices evaluated in a recent publication, providing a sleep detection performance of 0.93 (sensitivity) and wake detection performance of 0.60 (specificity) at a prediction probability threshold of 0.75. The conditional random field algorithm retained this performance for individual sleep parameters, including total sleep time, sleep efficiency, and wake after sleep onset (within the middle 50% to top 25% of the assessed devices). The only parameter where the model performance was lower was sleep onset latency (within the bottom 25% of all comparator devices). Using industry-best practices, we developed a sleep algorithm for use with the RW2 patch that can accurately detect sleep and wake windows compared to PSG-labeled sleep data. This algorithm may be used for a more complete understanding of well-being for patients with SMI in a real-world setting, without the need for PSG and a sleep lab.

Motion Cognitive Decoding of Cross-Subject Motor Imagery Guided on Different Visual Stimulus Materials.

Motor imagery (MI) plays an important role in brain-computer interfaces, especially in evoking event-related desynchronization and synchronization (ERD/S) rhythms in electroencephalogram (EEG) signals. However, the procedure for performing a MI task for a single subject is subjective, making it difficult to determine the actual situation of an individual's MI task and resulting in significant individual EEG response variations during motion cognitive decoding. To explore this issue, we designed three visual stimuli (arrow, human, and robot), each of which was used to present three MI tasks (left arm, right arm, and feet), and evaluated differences in brain response in terms of ERD/S rhythms. To compare subject-specific variations of different visual stimuli, a novel cross-subject MI-EEG classification method was proposed for the three visual stimuli. The proposed method employed a covariance matrix centroid alignment for preprocessing of EEG samples, followed by a model agnostic meta-learning method for cross-subject MI-EEG classification. The experimental results showed that robot stimulus materials were better than arrow or human stimulus materials, with an optimal cross-subject motion cognitive decoding accuracy of 79.04%. Moreover, the proposed method produced robust classification of cross-subject MI-EEG signal decoding, showing superior results to conventional methods on collected EEG signals.

DLBclass: A Probabilistic Molecular Classifier to Guide Clinical Investigation and Practice in DLBCL.

Diffuse large B-cell lymphoma (DLBCL) is a clinically and molecularly heterogeneous disease. The increasing recognition and targeting of genetically defined DLBCLs highlights the need for robust classification algorithms. We previously characterized recurrent genetic alterations in DLBCL and identified five discrete subtypes, Clusters 1-5 (C1-C5), with unique mechanisms of transformation, immune evasion, candidate treatment targets and different outcomes following standard first-line therapy. Herein, we validate the C1-C5 DLBCL taxonomy in an independent dataset and use the expanded series of 699 primary DLBCLs to develop a probabilistic molecular classifier and confirm its performance in an independent test set. Using our previously assigned cluster labels as a reference, we systematically compared multiple machine learning models and strategies for input feature dimensionality reduction with a newly developed performance metric that captured the relationship between accuracy and confidence of class assignments. The winning neural network model, DLBclass, assigned all cases in the training/validation and independent test sets with 91% and 89% accuracies, respectively. In the 75% of cases with confidence >0.7, DLBclass assignments were accurate in 97% of the training/validation set and 98% of the test set. DLBclass enables robust prospective classification of single cases for inclusion in genetically guided clinical trials or practice and represents a framework for the development of genomic-based classification methods in other cancers.

Mixed neuroendocrine non-neuroendocrine tumors: The quest for evidence

Mixed neuroendocrine non-neuroendocrine neoplasms (MiNENs) are rare mixed tumors containing both neuroendocrine and non-neuroendocrine components that occupy at least 30% of the whole tumor. Biologically, both components appear to derive from an identical cellular precursor undergoing early dual differentiation or late transdifferentiation. While our understanding of MiNENs has improved in recent years, many areas of uncertainty remain. In this context, setting diagnostic criteria capable of capturing the continuum of disease biology while providing clinically meaningful information in terms of prognosis and response to treatments appears vital to advance the field and improve patients’ outcomes. Evidence is needed to generate robust classification schemes, and multi-institutional cooperation will likely play a crucial role in building adequately powered cohorts to address some of the most pressing questions discussed in this Editorial. What is the minimum representation for each component needed to define MiNENs? How can the epidemiology of MiNENs change according to different diagnostic definitions? How can we generate the clinical evidence needed to optimize the management of MiNENs?

Leveraging Deep Learning for Robust Structural Damage Detection and Classification: A Transfer Learning Approach via CNN

Transfer learning techniques for structural health monitoring in bridge-type structures are investigated, focusing on model generalizability and domain adaptation challenges. Finite element models of bridge-type structures with varying geometry were simulated using the OpenSeesPy platform. Different levels of damage states were introduced at the midspans of these models, and Gaussian-based load time histories were applied at mid-span for dynamic time-history analysis to calculate acceleration data. Then, this acceleration time-history series was transformed into grayscale images, serving as inputs for a Convolutional Neural Network developed to detect and classify structural damage states. Initially, it was trained and tested on datasets derived from a Single-Source Domain structure, achieving perfect accuracy (1.0) in a ten-label multi-class classification task. However, this accuracy significantly decreased when the model was sequentially tested on structures with different geometry without retraining. To address this challenge, it is proposed that transfer learning be employed via feature extraction and joint training. The model showed a reduction in accuracy percentage when adapting from a Single-Source Domain to Multiple-Target Domains, revealing potential issues with non-homogeneous data distribution and catastrophic forgetting. Conversely, joint training, which involves training on all datasets except the specific Target Domain, generated a generalized network that effectively mitigated these issues and maintained high accuracy in predicting unseen class labels. This study highlights the integration of simulation data into the Deep Learning-based SHM framework, demonstrating that a generalized model created via Joint Learning utilizing FEM can potentially reduce the consequences of modeling errors and operational uncertainties unavoidable in real-world applications.

Robust Automatic Modulation Classification via a Lightweight Temporal Hybrid Neural Network.

In the rapidly developing field of wireless communications, the precise classification of modulated signals is essential for optimizing spectrum utilization and improving communication quality. However, existing networks face challenges in robustness against signals containing phase shift keying and computational efficiency. This paper introduces TCN-GRU, a lightweight model that combines the advantages of multiscale feature extraction of the temporal convolutional network (TCN) and global sequence modeling of gated recurrent unit (GRU). Compared to the state-of-the-art MCLDNN, TCN-GRU reduces parameters by 37.6%, achieving an accuracy of 0.6156 and 0.6466 on the RadioML2016.10a and RadioML2016.10b, respectively (versus MCLDNN's 0.6101 and 0.6462). Furthermore, TCN-GRU demonstrates superior ability in distinguishing challenging modulations such as QAM16 and QAM64, and it improves classification accuracy by about 10.5% compared to MCLDNN. These results suggest that TCN-GRU is a robust and efficient solution for enhancing AMC in complex and noisy environments.

Combination of Deep and Statistical Features of the Tissue of Pathology Images to Classify and Diagnose the Degree of Malignancy of Prostate Cancer.

Prostate cancer is one of the most prevalent male-specific diseases, where early and accurate diagnosis is essential for effective treatment and preventing disease progression. Assessing disease severity involves analyzing histological tissue samples, which are graded from 1 (healthy) to 5 (severely malignant) based on pathological features. However, traditional manual grading is labor-intensive and prone to variability. This study addresses the challenge of automating prostate cancer classification by proposing a novel histological grade analysis approach. The method integrates the gray-level co-occurrence matrix (GLCM) for extracting texture features with Haar wavelet modification to enhance feature quality. A convolutional neural network (CNN) is then employed for robust classification. The proposed method was evaluated using statistical and performance metrics, achieving an average accuracy of 97.3%, a precision of 98%, and an AUC of 0.95. These results underscore the effectiveness of the approach in accurately categorizing prostate tissue grades. This study demonstrates the potential of automated classification methods to support pathologists, enhance diagnostic precision, and improve clinical outcomes in prostate cancer care.

Detecting Aggression in Language: From Diverse Data to Robust Classifiers

The automatic detection of aggressive language is a difficult challenge. Currently, three datasets are available in Polish, enabling the training of machine learning models to recognise different types of linguistic aggression. In this paper, we address the issues of the transferability of knowledge between datasets and training a single model that works best on all types of aggression. Due to data imbalance, we experiment with two loss functions dedicated to training on imbalanced data: Weighted Cross-Entropy and Focal loss. Using the Polish language HerBERT model, we present the results of experiments in the Cross-dataset scenario and the model results using the combined data. Our results show that (1) combining diverse types of linguistic aggression during training leads to a better-performing classifier and (2) Weighted Cross-Entropy outperforms other tested loss functions.

A Comprehensive Approach to Bias Mitigation for Sentiment Analysis of Social Media Data

Sentiment analysis is a vital component of natural language processing (NLP), enabling the classification of text into positive, negative, or neutral sentiments. It is widely used in customer feedback analysis and social media monitoring but faces a significant challenge: bias. Biases, often introduced through imbalanced training datasets, can distort model predictions and result in unfair outcomes. To address this, we propose a bias-aware sentiment analysis framework leveraging Bias-BERT (Bidirectional Encoder Representations from Transformers), a customized classifier designed to balance accuracy and fairness. Our approach begins with adapting the Jigsaw Unintended Bias in Toxicity Classification dataset by converting toxicity scores into sentiment labels, making it suitable for sentiment analysis. This process includes data preparation steps like cleaning, tokenization, and feature extraction, all aimed at reducing bias. At the heart of our method is a novel loss function incorporating a bias-aware term based on the Kullback–Leibler (KL) divergence. This term guides the model toward fair predictions by penalizing biased outputs while maintaining robust classification performance. Ethical considerations are integral to our framework, ensuring the responsible deployment of AI models. This methodology highlights a pathway to equitable sentiment analysis by actively mitigating dataset biases and promoting fairness in NLP applications.

Innovative CNN approach for reliable chicken meat classification in the poultry industry

In response to the burgeoning need for advanced object recognition and classification, this research embarks on a journey harnessing the formidable capabilities of Convolutional Neural Networks (CNNs). The central aim of this study revolves around the precise identification and categorization of objects, with a specific focus on the critical task of distinguishing between fresh and spoiled chicken meat. This study's overarching objective is to craft a robust CNN-based classification model that excels in discriminating between objects. In the context of our research, we set out to create a model adept at distinguishing between fresh and rotten chicken meat. This endeavor holds immense potential in augmenting food safety and elevating quality control standards within the poultry industry. Our research methodology entails meticulous data collection, which includes acquiring high-resolution images of chicken meat. This meticulously curated dataset serves as the bedrock for both training and testing our CNN model. To optimize the model, we employ the 'adam' optimizer, while critical performance metrics, such as accuracy, precision, recall, and the F1-score, are methodically computed to evaluate the model's effectiveness. Our experimental findings unveil the remarkable success of our CNN model, with consistent accuracy, precision, and recall metrics all reaching an impressive pinnacle of 94%. These metrics underscore the model's excellence in the realm of object classification, with a particular emphasis on its proficiency in distinguishing between fresh and rotten chicken meat. In summation, our research concludes that the CNN model has exhibited exceptional prowess in the domains of object recognition and classification. The model's high accuracy signifies its precision in furnishing accurate predictions, while its elevated precision and recall values accentuate its effectiveness in differentiating between object classes. Consequently, the CNN model stands as a robust foundation for future strides in object classification technology. As we peer into the horizon of future research, myriad opportunities beckon. Our CNN model's applicability extends beyond chicken meat classification, inviting exploration across diverse domains. Furthermore, the model's refinement and adaptation for specific challenges represent an exciting avenue for future work, promising heightened performance across a broader spectrum of object recognition tasks.

Real-Time Radar Classification Based on Software-Defined Radio Platforms: Enhancing Processing Speed and Accuracy with Graphics Processing Unit Acceleration.

This paper presents a comprehensive evaluation of real-time radar classification using software-defined radio (SDR) platforms. The transition from analog to digital technologies, facilitated by SDR, has revolutionized radio systems, offering unprecedented flexibility and reconfigurability through software-based operations. This advancement complements the role of radar signal parameters, encapsulated in the pulse description words (PDWs), which play a pivotal role in electronic support measure (ESM) systems, enabling the detection and classification of threat radars. This study proposes an SDR-based radar classification system that achieves real-time operation with enhanced processing speed. Employing the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm as a robust classifier, the system harnesses Graphical Processing Unit (GPU) parallelization for efficient radio frequency (RF) parameter extraction. The experimental results highlight the efficiency of this approach, demonstrating a notable improvement in processing speed while operating at a sampling rate of up to 200 MSps and achieving an accuracy of 89.7% for real-time radar classification.

On the existence of solutions to adversarial training in multiclass classification

Abstract Adversarial training is a min-max optimization problem that is designed to construct robust classifiers against adversarial perturbations of data. We study three models of adversarial training in the multiclass agnostic-classifier setting. We prove the existence of Borel measurable robust classifiers in each model and provide a unified perspective of the adversarial training problem, expanding the connections with optimal transport initiated by the authors in their previous work [21]. In addition, we develop new connections between adversarial training in the multiclass setting and total variation regularization. As a corollary of our results, we provide an alternative proof of the existence of Borel measurable solutions to the agnostic adversarial training problem in the binary classification setting.

How Does Social Support Detected Automatically in Discussion Forums Relate to Online Learning Burnout? The Moderating Role of Students’ Self-Regulated Learning

Engaging students in online discussion forums with social support holds significant potential for preventing and alleviating student burnout. However, the mechanisms by which different types of social support influence learning burnout remain poorly understood. Additionally, existing methods for detecting social support detection are limited in both practical application and theoretical advancement. This study addresses these gaps by developing a robust text classification model for social support and examining its effects on online learning burnout among learners with varying levels of self-regulated learning. We first developed a robust natural language processing model based on bidirectional encoder representations from transformers - bidirectional long short-term memory (BERT-Bi-LSTM) framework, trained on 11226 manually labeled posts from various course forums. This model was then applied to classify forum posts from an educational technology course over one semester. Multiple regression analysis revealed that informational support was negatively associated with two dimensions of learning burnout: emotional exhaustion and improper behavior, and emotional support was negatively correlated with emotional exhaustion and a low sense of achievement. Moreover, a moderating effect analysis indicated that self-regulated learning moderated the negative associations between informational support and improper behavior, as well as between emotional support and emotional exhaustion, with stronger effects observed among learners with lower self-regulated learning. These findings contribute to advancing automated content analysis of social support and provide actionable insights for mitigating student burnout through targeted social support.