Breast cancer remains a leading cause of cancer-related deaths among women, highlighting the need for accurate computer-aided diagnosis systems (CADs). Convolutional neural networks (CNNs) have demonstrated substantial progress in medical image analysis, significantly improving diagnostic accuracy. This paper introduces Tri-ResNet, a triple-input model composed of parallel fine-tuned ResNet-based branches using transfer learning (TL) for efficient breast cancer classification. The model simultaneously processes full mammogram images (FMs), regions of interest images (ROIs), and contrastenhanced ROI images (CLAHE-enhanced ROIs) using Contrast-Limited Adaptive Histogram Equalization (CLAHE). Extensive experiments were conducted using multiple pre-trained models across single-input and multi-input architectures. Tri-ResNet achieved outstanding results on the Mini-DDSM, MIAS, and INbreast datasets, with peak performance on MIAS reaching 99.62% accuracy for normal{abnormal classification and 99.14% for benign{malignant classification, while maintaining competitive results on Mini-DDSM and INbreast. The model consistently outperformed single-input models and state-of-the-art approaches, demonstrating the effectiveness of multi-input CNNs for enhancing automated breast cancer diagnosis.

The graph coloring problem (GCP) is a classical NP-hard problem that aims to assign different colors to adjacent nodes while minimizing the total number of colors used. While previous studies have used graph neural networks (GNNs) to solve GCP, they rely only on randomly initialized node features or a trainable embedding layer, leaving other alternative node feature extraction methods unexplored. Therefore, this study explores eight node feature extraction methods, including positional and structural node features. We assess their impact on GNN performance for GCP and provide insights into why certain methods outperform others. Across 12 COLOR graphs and 3 large citation graphs, experimental results show that both the trainable embedding layer and node2vec achieve the strongest performance. Under different hyperparameter settings, embedding layer demonstrates consistent effectiveness in minimizing conflicts across GNN architectures, while node2vec demonstrates greater average performance and stability on large graphs. Compared to the embedding layer baseline, node2vec reduces mean conflicts by 43.63% and standard deviation by 27.23% on large citation graphs. However, their performance gains involve a computational trade-off: embedding layer requires backpropagation, and node2vec requires pre-computation for its biased random walks. Furthermore, positional node features give better prediction performance than structural ones, having approximately 9.3× lower mean conflicts in the best case.

The Internet of Health Things (IoHT) is a network of healthcare equipment, software, and systems that enable remote monitoring and healthcare services. Real-time health data is gathered via sensors. Even IoHT offers many benefits for modern smart healthcare, security concerns are increasing since IoHT devices lack appropriate processing power, storage capacity, and self-defense capabilities. In the healthcare sector, the use of machine learning (ML) for malware detection is vital for saving patients sensitive data. Therefore, it is essential to improve the accuracy and effectiveness of detection methods. ML models have been utilized to enhance the efficiency of malware detection. The main objectives of the attackers are to obtain personal information and take advantage of device flaws. Scientists are also devising diverse methods for identifying and analyzing malware to address these challenges. Given the continuous introduction of new malware by developers, it is highly tough to construct comprehensive algorithms for detecting such malware. Researchers have developed several ML and deep learning (DL) algorithms. The precision of these models will mainly contingent upon the amount of the training dataset. In addition, our work is divided into three primary stages: feature selection, prediction, and pre-processing. This work introduces feature selection technique that integrates two approaches, the first one Pearson correlation, to assess the correlation between features and identify significant features and Embedded method. These selected features are subsequently utilized in a classification model. Our method utilizes a soft voting classifier that combines multiple machine learning models (decision tree, logistic regression, gradient boost, random forest, and support vector machine) to detect malware. This approach creates a single model that incorporates the strengths of the combined models, resulting in the highest prediction accuracy. The proposed methodology surpasses previous research by reaching a 99.6 % accuracy rate, an F1 score of 0.9972 % a recall rate of 0.9998, a precision rate of 0.9947.

This study investigates the predictive power of learning management system (LMS) log data for student performance in higher education. Analyzing interactions from 114 students in a sports pedagogy course, we compared linear regression (LR), random forest regression (RFR), and support vector regression (SVR), each paired with mutual information (MI) and backward elimination (BE) feature selection. Results show LMS log data alone can effectively predict final grades, with SVR[Formula: see text]BE performing best ([Formula: see text], MAE [Formula: see text] 4.54). Feature selection, particularly BE, consistently improved model performance across all algorithms. Key findings include: LMS interactions strongly predict academic performance; SVR outperforms other algorithms in capturing complex educational data relationships; and BE’s superiority highlights the importance of feature interactions. This research advances educational data mining (EDM) by identifying optimal modeling approaches for LMS data, contributing to the development of early warning systems in online and blended learning environments.

Phishing is a common malicious cybercrime in which attackers trick people into revealing their sensitive information. Phishing attacks are serious, as they permit the attacker to perceive and steal the victim’s personal and sensitive information, such as credit card numbers and passwords, in a scam way while the victim is browsing the phishing website. Security experts are responsible for developing and continually improving their algorithms to save the community and guard its information. Traditional security techniques fail to do this effectively; experts are looking for new and robust methods. In this research, we use machine learning (ML) to improve phishing detection tasks. This technology efficiently influences the recognition of hidden patterns in large data inputs. The CMLM algorithm, used for detecting phishing website attacks, suffers from a high rate of indiscernibility relations, which reflect the lack of understanding of the outputs. Additionally, it does not account for imbalanced data. This paper proposes two new versions of the CMLM algorithm that effectively address these issues by integrating methods such as K-means clustering, Stability-correlation and correlation (ScC), rough set (RS) theory, principal component analysis (PCA), decision tree (DT), and deep learning (DL) in a controlled manner. The results show that the proposed methods demonstrate higher accuracy in detecting phishing websites than CMLM, achieving accuracies of 100%, 99.96%, and 99.81% across three key datasets. Compared to CMLM, the improvement margins are 0.47%, 3.08%, and 3.45% for DS1, DS2, and DS3, respectively.

Autonomous Unmanned Aerial Vehicle (UAV) navigation in dynamic and partially observable environments poses significant challenges, including real-time decision-making and robust obstacle avoidance. Traditional methods often struggle with adaptability, necessitating more advanced approaches. In this work, we propose a Deep Reinforcement Learning (DRL) framework for trajectory optimization, leveraging the Advantage Actor–Critic (A2C) algorithm. We further enhance stability, learning speed, and generalization by employing automatic hyperparameter tuning with Optuna. The proposed system is validated in a Software-in-the-Loop (SITL) simulation using AirSim, ensuring realistic flight dynamics and sensor feedback. Multi-modal observations — combining depth images, GPS, and target localization — improve situational awareness in partially observable conditions. Experimental results show that A2C tuned with Optuna boosts trajectory efficiency by 35.7%, reduces the collision rate to 0.97% and achieves a 74% success rate, while cutting training time by 42%. These findings confirm the effectiveness of using automated hyperparameter tuning for UAV motion planning and pave the way for real-world deployment of DRL-based UAV control systems. Furthermore, our study provides an in-depth comparison of training efficiency, convergence properties, and robustness across different algorithms, establishing a strong foundation for autonomous UAV navigation in challenging environments.

Videos have become a common medium for customers to share their feedback, driven by the widespread use of the internet and the resulting proliferation of diverse content across social media platforms. These videos often contain multimodal forms of data, such as text, audio, and visual components, making them rich sources for sentiment analysis (SA). Multimodal sentiment analysis (MSA) combines data from all these modalities to improve understanding of sentiments and achieve an accurate prediction. However, existing multimodal fusion methods do not perform satisfactorily well, as they either concatenate raw features or often overlook intricate interdependencies between modalities, failing to resolve conflicts between them. This work attempts to implement a novel fusion method, Hierarchical Attention-Based Fusion (HABF), which uses self-attention and cross-attention mechanisms in a hierarchical manner that prioritizes and integrates efficient features coming from all modalities. HABF combines unimodal features by assigning contextual weights to each modality, ensuring the most accurate representation of sentiments in videos. The support vector machine then classifies the fused representation into positive, neutral, or negative sentiments. The model is tested on two datasets: (1) the Customer-Generated Sentiment Videos (CGSV) dataset, which is restaurant review-based, and (2) the standard dataset, CMU-MOSI. The proposed model uses Bidirectional Long Short-Term Memory (BiLSTM) for text, Librosa for audio, and a Convolutional Neural Network (CNN) for visual data for feature extraction. The evaluated results are compared with existing systems using CMU-MOSI, achieving better accuracy with 82.43%. The proposed model shows an accuracy of 78.10% using the CGSV dataset. This study focuses mainly on improving MSA for customer satisfaction, dining experience, and service quality, and subsequently increasing better business strategies for restaurants.

Predictive cancer epidemiology applies machine learning to analyze historical medical data for early diagnosis, reduce cancer spread, and identify high-risk individuals. However, data quality issues and class imbalance often hinder these tasks. Existing class imbalance solutions are prone to information loss, weak recognition of the majority class, increased false positives, and reduced reliability. To overcome these drawbacks, we propose a novel enhanced boosting-based method, JEUBoost. First, JEUBoost reduces the majority class size by using Gaussian probability density to estimate sample probabilities and information entropy to measure sample informativeness, creating a more balanced dataset. Second, a modified metaheuristic algorithm, Jaya, is employed to improve probability estimation of high-quality samples by adjusting relevant parameters of the Gaussian model. Third, a customized cost function for Jaya is formulated as an optimization problem to minimize the model’s error rate. Experimental results demonstrate that the performance metrics, including Accuracy, Precision, Recall, F1-score, G-means, and Precision–Recall curves, achieved by JEUBoost range from 68% to 99%. Compared to conventional class imbalance methods, JEUBoost improved Precision by 2.6%, Recall by 5.5%, G-means by 3.2%, and average Precision of the curve by 8.7%, while reducing variance by 82.71%, demonstrating consistent performance gains across all key metrics.

Incorporating the features of genetic algorithm (GA) and particle swarm optimization (PSO), a new algorithm named PSO-GA is designed for single-objective continuous optimization problems (SOCOPs). Additional perturbation rules have been integrated into PSO to develop an enhanced heuristic named multi-rule PSO (MRPSO) for the same. Likewise, grey wolf optimizer (GWO) is modified by absorbing two more perturbation rules and is named multi-rule GWO (MRGWO). Also, GA and MRGWO are amalgamated to develop a new algorithm named GWO-GA. Finally, the merits of the two sets of heuristic approaches — {PSO, MRPSO, GA} and {GWO, EGWO, MRGWO} — are exploited using Q-learning to develop a hyper-heuristic, named PSO-GWO-Q, for the global optimization to SOCOPs, where the algorithm enhanced GWO (EGWO) is taken from the literature. All the algorithms have been tested against 50 benchmark test problems and it is observed that only PSO-GWO-Q provides results with a desired precision for the studied test problems. Comparing the consistency and efficiency of PSO-GWO-Q with some state-of-the-art algorithms for the SOCOPs using standard statistical tests, it is observed that the designed hyper-heuristic outperforms the others.

This paper presents a study investigating the optimization of well-known NLP algorithms and approaches for the Georgian language, known for its unique linguistic features. Standard methods effective for well-resourced languages, including pretrained models like mBERT and embedding methods such as FastText, may lack flexibility and efficiency when applied to Georgian, often resulting in increased complexity and effort. To address these challenges, we propose a novel approach that leverages Georgian’s rich morphology, including case inflections, extensive suffixation, verb agreement, and conjugation patterns. This method refines algorithms such as Minimum Editing Distance, Text Classification, Language Modeling, and word-level semantic similarity by incorporating language-specific characteristics. Our approach reduces data sparsity and model complexity while preserving accuracy. Although developed for Georgian, it is also relevant for other fusional and agglutinative languages and contributes to reducing dependence on large corpora, supporting the creation of more human-like text.