Accurate differentiation of mononuclear cells is essential for diagnosing hematological conditions. Nonetheless, their similar morphology can lead to misclassification. The application of artificial intelligence enables the extraction of discriminative features from cell images through advanced mathematical modeling. While a compact model is necessary to enable dynamic-monitoring in clinical practice, achieving high accuracy typically requires large, well-balanced datasets, presenting significant challenges in balancing model complexity and classification performance. This study compares traditional features like Local Binary Patterns (LBP), Histogram of Oriented Gradients (HOG), and Gray-Level Co-occurrence Matrix (GLCM) with transfer learning using fine-tuned Vision transformer (ViT) for capturing features of images. Using a dataset of 4,315 images focusing on mononuclear cells, we integrated class balancing techniques and tested models with ten classifiers. Results showed HOG with XGBoost achieved 0.77 accuracy, while fine-tuned ViT with augmentation and XGBoost reached 0.82. The computational intensity of ViT is offset by its real-time monitoring potential.

Protein–protein interactions (PPIs) are essential to understand biological processes and cellular functions. Experimental methods for identifying PPIs are often time-consuming and costly, motivating the development of accurate computational prediction approaches. In this paper, we combine embeddings from the state-of-the-art protein language model ESM-2 with handcrafted descriptors including short-linear motifs representation based on the ELM database. The integration of both deep learning-based and biologically meaningful handcrafted features aims to capture global and local characteristics of protein sequences. These extracted features are subsequently fused and used to train machine learning classifiers to determine interaction likelihoods between protein pairs. Experimental results on benchmark PPI datasets demonstrate that our hybrid feature strategy significantly improves prediction performance, highlighting the complementary strengths of learned embeddings and handcrafted descriptors. This work provides a robust and extensible framework for sequence-based PPI prediction, paving the way for scalable and accurate interactome mapping.

Emotion detection on social media, especially related to mental health, has emerged as a significant topic in natural language processing (NLP). This study fine-tunes Transformer-based models — IndoBERT, RoBERTa, and DistilBERT — to identify emotions linked to depression in Indonesian tweets. A dataset of 5271 tweets was compiled using domain-specific keywords derived from mental health literature and expert inputs. Preprocessing follows standard NLP practices to ensure linguistic consistency and minimize noise. Several feature extraction techniques, including TF-IDF, Bag of Words (BoW), and Word2Vec, were evaluated using traditional machine learning (ML) classifiers under a 10-fold cross-validation scheme. The best performance was achieved by DistilBERT combined with TF-IDF and support vector machine (SVM), yielding an F1-score of 0.699 and accuracy of 0.705. Although the results remain moderate compared to high-resource languages, they mark a meaningful step toward modeling emotional indicators of depression in Indonesian social media contexts. The study underscores the importance of domain-specific and culturally aware approaches in advancing NLP applications for mental health.

Malware threatens cybersecurity by enabling data theft, unauthorized access, and extortion. Traditional malware detection systems (MDS) struggle with the increasing volume and complexity of malware. While machine learning (ML) and deep learning (DL) offer promising solutions, they remain vulnerable to adversarial attacks that evade detection. Recent research focuses on developing adversarial datasets to retrain ML/DL-based malware detection systems, enhancing their robustness against adversarial attacks. While these methods improve detection of adversarial samples, they also cause more misclassification of non-adversarial data due to overfitting. These methods lack scalability when ML/DL-based MDS are retrained in isolation, without utilizing knowledge from other MDS with retrained models, leading to inefficiency and waste. To tackle these issues, we introduce ProDef-MDS, a proactive defense system that integrates an Adversarial Restoration (AR) module to mitigate adversarial perturbations and recover inputs to a correctly classifiable form before passing them into the malware classification model. We focus on portable executable (PE) malware within Windows OS to evaluate our approach’s effectiveness across various scenarios, including those with adversarial data generated from five white-box attacks, including Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD)10, PGD100, DeepFool, Carlini and Wagner (CW) and one black-box attack Auxiliary Classifier Generative Adversarial Networks (ACGAN). Additionally, we assess our approach with non-adversarial data to demonstrate its effectiveness in adversarial detection without compromising non-adversarial performance. The results obtained from the real-world dataset indicate enhanced robustness and minimal overhead, offering a proactive solution to adversarial threats in MDS. This approach outperforms retraining defense method in five white-box attacks and also indicates better performance in non-adversarial scenarios.

In the field of natural language processing (NLP), sentiment analysis is very important for comprehending user views and opinions. By using the Kaggle ChatGPT sentiment analysis dataset, this work investigates the use of sophisticated ensemble machine learning algorithms to enhance ChatGPT’s sentiment classification performance. Due to its skewed distribution, the dataset poses particular difficulties in reaching high performance metrics. For sentiment classification, we evaluate the performance of many boosting techniques such as AdaBoost, Gradient Boosting Machine (GBM), XGBoost, LightGBM (LGBM), and CatBoost. Moreover, ensemble techniques such as majority voting and stacking are used to improve classification results. While majority voting combines predictions for robust classification, stacking specifically uses many base learners in conjunction with a meta-model to maximize performance. With a classification accuracy of 88.57%, precision of 88.67%, recall of 88.57%, and an F-measure of 88.57%, our tests show that stacking performs better than any other technique. Its efficient mistake mitigation is further demonstrated by its False Positive Rate (FPR) of 0.09 and False Negative Rate (FNR) of 0.13. These findings highlight how ensemble methods — in particular, stacking — can be used to handle skewed data and enhance sentiment classification capabilities. In light of ChatGPT and related conversational artificial intelligence (AI) technologies, the results offer important new information for the creation of trustworthy sentiment analysis systems.

Early recognition of plant diseases is crucial, and one practical approach is using deep learning models. Models, such as MLP-Mixer and gMLP, based on multi-layer perceptron, offer a compelling option due to their simple architecture. This study aimed to assess the performance of these models in classifying crop leaf diseases. Our findings reveal that both MLP-Mixer and gMLP models, with their similar architectures, exhibit promising performance. We conducted tests using public potato and wheat datasets to evaluate their classification performance. Furthermore, we incorporated gradient centralization during training to enhance the models’ generalization performance. The results indicate that both MLP-Mixer and gMLP achieved classification performance above 0.9100 for both datasets. Specifically, MLP-Mixer achieved 0.9819, and gMLP achieved 0.9873 for potato leaf diseases, while for wheat leaf diseases, MLP-Mixer achieved 0.9121, and gMLP achieved 0.9189. These outcomes emphasize the potential of these models in classifying and identifying diseases in crop leaves.

Today, due to the problem of environmental pollution, water, and other factors have caused many dangerous diseases, including cancer. According to recent statistics, breast cancer is one of the leading diseases in women, and this disease tends to increase more and more. To detect and diagnose the disease, doctors perform many examinations: self-examination, clinical examination, X-ray, ultrasound screening, etc., in which X-ray is a highly effective method. This study proposes an approach to detecting and classifying breast cancer on an X-ray dataset using a refined Vision Transformer (ViT), ViT-B32. The considered dataset contains about 7000 X-ray images from patients aged 27 to 90, labeled as malignant, benign, or normal. As presented in scenarios, the study yielded positive results, with 91% to 94% in ACC and F1-score metrics. Furthermore, it has shown that the results obtained for breast cancer detection on X-ray images using the fine-tuned ViT architecture outperformed CNN models such as VGG16, MobileNet, Xception, ResNet50, and some state-of-the-art approaches.

Mental health (MH) assessment and prediction have become critical areas of focus in healthcare, leveraging developments in natural language processing (NLP). Recent advancements in machine learning have facilitated the exploration of predictive models for MH based on user-generated comments that overlooked the integration of emotional attention mechanisms. The methods often struggle with contextual nuances and emotional subtleties, leading to suboptimal predictions. The prevailing challenge lies in accurately understanding the emotional context embedded within textual comments, which is crucial for effective prediction and intervention. In this research, we introduce a novel approach employing contextual emotional transformer-based models (CETM) for comment analysis in MH case prediction. CETM leverages state-of-the-art transformer architectures enhanced with contextual embedding layers and emotional attention mechanisms for MH case prediction. By incorporating contextual information and emotional cues, CETM captures the underlying emotional states and MH indicators expressed in user comments. Through extensive experimentation and evaluation, both Roberta and bidirectional encoder representations from transformers (BERT) models exhibited enhanced accuracy, precision, recall, and F1 scores compared to their counterparts lacking emotional attention. Notably, the Roberta model attained a greater accuracy of 94.5% when matched to BERT’s 87.6% when emotional attention was employed. Hence, by incorporating emotional context into the predictive model, we achieved significant improvements, which offers promising avenues for more precise and personalized MH interventions.