Heterogeneous ensemble learning and improved binary escape algorithms for computer-aided detection of breast cancer

BALSO-DTV: Binary artificial locust swarm optimization algorithm boosted with agent motion-based dynamic time-varying S-shaped transfer functions for feature selection in high-dimensional data

A Novel Personal Best- Control Binary Particle Swarm Optimization (NPbest-BPSO) based electromyography (EMG) signal feature selection and classification

Metaheuristic algorithms offer advantages over traditional methods in handling complex feature dependencies and avoiding local optima, making them increasingly favored in feature selection tasks.The Artificial Algae Algorithm (AAA), a metaheuristic inspired by the behavior of natural algae populations, has demonstrated rapid convergence and efficient feature selection in low-dimensional problems.However, in high-dimensional feature selection tasks, the algorithm often suffers from premature convergence, leading to entrapment in local optima.To address this issue, this paper proposes a ReliefF-Guided Binary Artificial Algae Algorithm (RG-BAAA) specifically designed for high-dimensional feature selection, which effectively guides the search process to overcome premature convergence.Specifically, the ReliefF algorithm is first used to assess feature importance, and the evaluation results are leveraged to guide the initialization of the population, ensuring that the algorithm focuses on key feature regions from the outset, thereby avoiding inefficient searches caused by random initialization.Next, the feature evaluation scores drive a weighted dynamic algae cell selection strategy, enhancing the directional focus of the search process and helping the algorithm quickly avoid local optima, reducing the risk of premature convergence.Additionally, an adaptive energy loss strategy is introduced to balance the exploration and exploitation phases, further improving the search direction and convergence performance.The experimental results using the random forest (the most effective classifier) show that RG-BAAA achieved the highest accuracy rate on all ten high-dimensional biomedical datasets.On the 9Tumors dataset, its accuracy rate was as high as 0.8324, surpassing the accuracy rates of the top ten most advanced meta-heuristic benchmark methods recently.

This paper addresses the challenge of enhancing the efficiency of studying binary arithmetic algorithms within the framework of training specialists in the field of computer technology. Purpose: the goal is to develop an educational web service designed for a comprehensive study of binary arithmetic algorithms. Methods: the methodological foundation is based on the analysis of existing software solutions and the design of a modular architecture for the web application. Results: unlike existing counterparts, the developed service implements a complete cycle of transformations and calculations, ranging from converting numbers into direct, reverse, and complement codes to executing basic arithmetic operations with support for variable computational methods. A key advantage of this solution is the module for generating detailed step-by-step solutions, which facilitates a deeper understanding of algorithmic principles. The scientific novelty of the work lies in the creation of a unified software platform that combines variable methods of binary computation with algorithmically transparent visualization of each stage of data processing. Practical significance: the results of a comparative analysis affirm the feasibility of using the developed service in educational practice. The comprehensive nature of the implemented functions and the high degree of detail in the computational processes create prerequisites for its application in the study of disciplines related to computer architecture and the theoretical foundations of computer science.

Background: Facial expressions of pain are essential for pain assessment, yet subjective pain reports often vary between sexes. Traditional self-report measures are prone to bias, and objective methods are needed for more reliable pain evaluation. Objective: To develop and validate a subjectivity-free automated tool to assess acute low back pain using facial expressions recorded during a functional spinal extension task. Participants: Thirty healthy adults, aged 18–40 years. Methods: Participants received intramuscular injections of hypertonic (pain) and isotonic (placebo) saline in the lumbar region during separate sessions. Facial expressions were video-recorded during a submaximal lumbar extension task and analyzed using a custom software based on Haar Cascade and Local Binary Pattern Histogram algorithms, which are techniques that do not require neither training data nor subjective labeling, contrary to what happens in deep learning solutions. Results: The tool successfully detected significant differences in facial expressions between pain, placebo, and pain-free conditions (p < 0.001). Test–retest reliability was good (ICC = 0.85). While both sexes showed similar facial expression patterns during pain, males reported higher pain scores on the numeric rating scale (p < 0.01). Pain significantly reduced steadiness of force in both sexes. Conclusion: The automated tool objectively quantified facial expressions associated with acute low back pain and revealed sex-related differences in subjective pain perception. This multimodal approach integrating expression analysis, physical performance, and self-report may enhance the accuracy of pain assessment in physiotherapy settings.

Deep learning models often encounter two key challenges in developing intelligent and scalable forecasting frameworks for renewable energy systems: input feature space dimensionality and sensitivity to hyperparameter settings. These limitations increase computational cost and compromise generalization and robustness. This paper presents a hybrid deep learning-optimization framework that leverages cognitively inspired metaheuristics to address these challenges, employing the Binary iHow Optimization Algorithm (biHOW) for feature selection and its continuous counterpart, iHOW, for hyperparameter tuning. Both variants emulate human cognitive phases-data absorption, information analysis, reinstitution, and adaptive knowledge development enabling efficient traversal of complex search spaces. Using the Multi-Scale Attention Network (MSAN) as the forecasting backbone, which is well suited for modeling renewable energy time series due to its ability to capture multi-scale temporal dependencies ranging from short-term fluctuations to long-term seasonal patterns, the proposed framework achieved high accuracy for wind and solar generation prediction. The MSAN model attained Mean Squared Errors (MSE) of 0.0105 for wind and 0.0976 for solar forecasting. Applying biHOW for feature selection reduced the average misclassification rate to 0.3925 (wind) and 0.4161 (solar) while identifying compact, interpretable feature subsets. The iHOW optimizer further fine-tuned architectural and training parameters, decreasing MSE to [Formula: see text] for wind and [Formula: see text] for solar, outperforming state-of-the-art metaheuristics including HHO, GWO, PSO, and JAYA. These findings demonstrate the effectiveness of iHOW-based optimization in enhancing forecasting accuracy and computational scalability. The proposed hybrid framework supports adaptive forecasting for intelligent energy management within modern smart grids.

The Set Covering Problem is a fundamental NP-hard problem in combinatorial optimization and plays a central role in a wide range of industrial decision-making processes, including logistics planning, scheduling, facility location, network design, and resource allocation. In many real-world contexts, problems of this type are large in scale and highly constrained, which makes exact solution methods computationally impractical and encourages the use of metaheuristic approaches capable of producing high-quality solutions within limited time budgets. In this work, we propose a discrete adaptation of the Dream Optimization Algorithm, focusing on the challenges that emerge when algorithms originally designed for continuous search spaces are applied to binary and strongly constrained models. The continuous search process is mapped onto the binary decision space through a fixed discretization scheme. As a consequence of this transformation, some constraints may not be met, underscoring the importance of effective feasibility restoration mechanisms. Because the discretization stage may produce infeasible solutions and frequently induces plateaus that hinder further improvement, an explicit repair phase becomes necessary to restore feasibility and promote effective search progression. To strengthen this process, the study introduces an adaptive control mechanism based on bandit driven operator selection, which dynamically chooses among different repair procedures during the search. Experimental results on benchmark instances show that the proposed approach consistently achieves high quality solutions with low relative deviation from known optima and stable behavior across independent runs.

The article was submitted to be part of a guest-edited issue. An investigation by the publisher found a number of articles, including this one, with a number of concerns, including but not limited to compromised editorial handling and peer review process, inappropriate or irrelevant references or not being in scope of the journal or guest-edited issue. Based on the investigation's findings the publisher, in consultation with the Editor-in-Chief therefore no longer has confidence in the results and conclusions of this article. S. Priya has not explicitly stated whether she agrees or disagrees with the retraction. K. Pradeep Mohan Kumar has not responded to correspondence regarding the retraction.

Abstract This research introduces an advanced diagnostic framework founded on dissolved gas analysis (DGA) for enhancing both the precision and efficiency of fault detection in power transformers. The framework is designed to accurately differentiate among six predefined fault categories by employing the bat algorithm to optimize the hyperparameters of a support vector machine trained on DGA-derived data. To strengthen the algorithm’s exploration capacity and to minimize the risk of premature convergence, chaotic dynamics were embedded within the optimization process, yielding a more balanced search between global and local regions of the solution space. Furthermore, the proposed methodology integrates the generation of multiple feature representations, which are subsequently refined through dimensionality reduction using an improved binary bat algorithm capable of selecting the most informative features. This dual-stage optimization process significantly enhances the classification accuracy while concurrently lowering computational demands. To further address the inherent issue of data imbalance in the training set, the synthetic minority over-sampling technique was employed, contributing to greater model robustness and improved generalization capability. The experimental findings confirm that the proposed hybrid diagnostic framework surpasses several state-of-the-art and machine learning-based approaches in terms of accuracy, computational efficiency, and stability, establishing it as a promising tool for reliable transformer fault classification and condition monitoring.

Comparative analysis for classification of cervical cancer cells using binary red deer algorithm and ResNet framework

Multi-objective neural architecture search combining binary artificial bee colony algorithm for dynamic hand gesture recognition

The automatic diagnosis of lung cancer using chest X-ray (CXR) images has significantly advanced with progress in computing, machine learning, and deep learning. However, detecting lesions and nodules remains challenging due to CXR limitations. Early lung cancer detection is critical for successful treatment, but current AI algorithms often rely on large annotated datasets, which are not always available. To address this, a novel multi-classification deep learning framework is proposed that combines CXR and CT images. This approach leverages the detailed feature detection capabilities of CT scans alongside the complementary views from CXRs, improving early-stage lung cancer detection and classification precision. The proposed framework integrates the Binary Adaptive Crossover Particle Whale Optimization Algorithm-S (BACP-WOA-S) to efficiently address high-dimensionality challenges. The proposed framework includes key contributions: Employing a Binary Whale Optimization Algorithm variation (BWOA) is used in the feature selection (FS) layer to reduce computational complexity while preserving key features. Additionally, a Feed-Forward Neural Network (FFNN) is utilized in the Deep Learning (DL) layer to optimize characteristics like layers, neurons, and activation functions. The framework, powered by H2O for scalable deep learning processes, achieves a 97.6% accuracy rate, showcasing its effectiveness in early lung cancer detection.

The Arctic Puffin Optimization (APO) Algorithm is a recently proposed metaheuristic algorithm that has been widely applied to solve optimization problems in continuous spaces. However, it cannot be directly used to solve combinatorial optimization problems in discrete spaces. To address these limitations, a Binary Arctic Puffin Optimization (BAPO) Algorithm is proposed, focusing on developing transfer functions to convert the algorithm’s continuous solutions into discrete binary solutions. Two primary transfer function types, S-shaped and V-shaped, are commonly employed. Experimental analysis identifies optimal functions for different algorithmic stages. These are then integrated with a conversion factor to propose a hybrid transfer function for the binarization of the Puffin Optimization Algorithm. To address the issue of slow particle convergence in the later stages of the exploration phase and the tendency to overlook high-quality solutions during the exploitation phase in the binary algorithm, logarithmic inertia weight and the golden sine strategy are incorporated, respectively, for improvement. Simulation experiments were conducted to solve both single-dimensional and multidimensional 0–1 knapsack problems. Experimental data and convergence curves, including mean values and standard deviations, were analyzed. The results demonstrate that the binary Arctic puffin optimization algorithm exhibits excellent convergence, stability, and fast search speed.

Embedded detection and recognition systems for surveillance video are in urgent demand in the security field. However, traditional methods face limitations, including poor real-time performance, high resource consumption, and limited generalization in complex scenarios. To this end, this study proposes a dual-engine embedded face detection and recognition framework that optimizes performance by synergistically integrating YOLO v5Face with attention-enhanced Faster Regions with Convolutional Neural Network. The system adopts a dual engine cascade architecture: YOLO5Face is responsible for fast initial face screening, while Faster Regions with Convolutional Neural Network, which integrates spatial and channel attention mechanisms, accurately recognizes key targets. By synergistically optimizing speed and accuracy through feature reuse and structural fusion techniques, and by combining the feature-extraction capabilities of the local binary pattern histogram algorithm based on hierarchical feature pyramids, a dynamic background suppression module is used to reduce false positives in complex scenes. The experimental results on the WIDER FACE and Face Detection Data Set and Benchmark datasets show that the accuracy of our system reaches 99.1%, with a loss rate as low as 0.08, significantly better than the comparison systems Visual Transformer Convolutional Neural Network Fusion (accuracy 98.16±0.23%) and Additive Marginal Soft Maximum Loss Convolutional Multi-scale Transformer (accuracy 97.42±0.34%); The system converges to a loss of less than 0.1 within 200 iterations, with a response time of only 28 ms, much faster than the fusion of Visual Transformer Convolutional Neural Network (78-85 ms). The above results show that the proposed method effectively addresses the problems of poor real-time performance, resource constraints, and insufficient scene generalization, offering efficient, lightweight new ideas for system development and promoting the intelligent and efficient development of security terminals.

Cancer microarray datasets are distinguished by their high dimensionality and a relatively small sample sizes, which presents significant challenges for accurate cancer classification. Gene selection therefore becomes essential to eliminate irrelevant genes and improve classification accuracy. This paper presents a hybrid approach combining filter and wrapper techniques for gene selection, integrating an improved binary firefly algorithm and the support vector machine classifier. The objective is to select the most cancer-related genes to decrease computation time and enhance classification model performance. Three filter methods (Information Gain Ratio, ReliefF, and Correlation-based Feature Selection) are used in ensemble with the enhanced binary firefly algorithm. The firefly algorithm’s exploration and exploitation capabilities are improved through opposition-based learning during initialization and movement of the fireflies. Additionally, a mutation step is added to improve the diversification of fireflies. To validate our approach, we conducted an experimental study on eight public benchmark datasets and compared it to several recent gene selection methods used for cancer gene expression data classification. The results reveal that the suggested methodology enhances classifier performance while reducing data volume by finding a limited group of genes with strong predictive power for cancer classification.

Cardiovascular disease (CVD) is a prevalent global health issue and one of the leading causes of death. Aging and air pollution are well-established risk factors for CVD. This study aims to investigate the association between air pollutants (sulfur dioxide, carbon monoxide, PM1, PM2.5, nitrogen dioxide, ozone) and the risk of heart disease. Utilizing data from the China Health and Retirement Longitudinal Study (CHARLS) and the China High Air Pollutants (CHAP) database, we employed multivariable-adjusted logistic regression to analyze the relationship between pollutants and heart disease. Additionally, six binary classification machine learning algorithms—AdaBoost, Decision Tree, LightGBM, XGBoost, Random Forest, and GBDT—were used to construct predictive models. The models incorporated air pollutant concentrations (SO₂, CO, PM1, etc.) as core features, along with covariates such as gender, age, and hypertension. The data were split into an 80% training set and a 20% test set, with cross-validation applied to ensure robustness. Multivariable regression analysis revealed that after adjusting for multiple covariates (including BMI, blood glucose, and other pollutants), each 1-unit increase in SO₂ concentration was associated with an odds ratio (OR) of 1.040 for heart disease (95% confidence interval [CI]: 1.027–1.054, p < 0.00001). Among the machine learning models, Random Forest exhibited the best performance, with an AUC of 0.794 in the training set and 0.726 in the test set. SHAP analysis confirmed that SO₂ was the most impactful pollutant. Subgroup analysis indicated a significant interaction between SO₂ and household registration type (p < 0.05). Future research should further explore the mechanisms underlying SO₂-induced cardiac damage and optimize the applicability of predictive models.

Lipschitz-based classification provides a flexible framework for general metric spaces, naturally adapting to complex data structures without assuming linearity. However, direct applications of classical extensions often yield decision boundaries equivalent to the 1-Nearest Neighbour classifier, leading to overfitting and sensitivity to noise. Addressing this limitation, this paper introduces a novel binary classification algorithm that integrates probabilistic kernel smoothing with explicit Lipschitz extensions. We approximate the conditional probability of class membership by extending smoothed labels through a family of bounded Lipschitz functions. Theoretically, we prove that while direct extensions of binary labels collapse to nearest-neighbour rules, our probabilistic approach guarantees controlled complexity and stability. Experimentally, evaluations on synthetic and real-world datasets demonstrate that this methodology generates smooth, interpretable decision boundaries resilient to outliers. The results confirm that combining kernel smoothing with adaptive Lipschitz extensions yields performance competitive with state-of-the-art methods while offering superior geometric interpretability.

The high mortality rate of metastatic cutaneous melanoma (SKCM) remains a major challenge in clinical treatment. This study used single-cell RNA sequencing (scRNA-Seq) technology to compare the differences between metastatic and primary tumour cells. By manually annotating cell types, significant disparities in cell communication patterns and functional pathways between the two groups were identified. Combined with transcriptomic data, differential gene analysis was performed to screen out a core gene set associated with tumour metastasis. To achieve accurate prediction of tumour metastasis, this study innovatively constructed a binary classification algorithm (PSO-SVM) integrating particle swarm optimisation (PSO) and support vector machines (SVMs). This model optimises SVM parameters via the PSO algorithm, addressing the limitations of traditional machine learning models such as insufficient accuracy and poor generalization ability in tumour metastasis prediction. Verified by comparison with mainstream machine learning methods, the PSO-SVM model exhibited superior classification performance and successfully identified five key metastasis-related genes: SFN, S100A8, KLF5, ARL4D and TINCR. Furthermore, the expression differences of these genes in the metastatic group were verified at the single-cell level, clarifying their regulatory roles in different cell types and states. Through an innovative analytical strategy integrating single-cell and transcriptomic data, this study elucidated the core molecular mechanisms of SKCM metastasis and key regulatory pathways in the tumour microenvironment, providing potential biomarkers and therapeutic targets for the early diagnosis and targeted treatment of SKCM metastasis. This PSO-SVM-integrated analysis method also offers new insights for research on metastasis mechanisms of other cancers.

To develop machine learning (ML) models to predict the probability at baseline of achieving low disease activity (LDA) and high health-related quality of life (HRQOL) in patients with psoriatic arthritis (PsA) or axial spondyloarthritis (axSpA) treated with secukinumab (SEC). AQUILA is an ongoing multicenter, prospective, noninterventional study assessing the effectiveness and safety of SEC in patients with active PsA or axSpA in Germany. Data from 1961 participants were used to develop ML models for predicting treatment outcomes. We investigated baseline prediction of achieving LDA and high HRQOL at week 16 using binary ML algorithms, identifying main predictors for LDA and high HRQOL and their direction of influence. In addition, explainable artificial intelligence (XAI) estimated the importance and impact of each predictor based on how it affected the change in individual patient predictions. In PsA, the main LDA predictors were patient global assessment, physician global assessment, pretreatment with biologic disease-modifying antirheumatic drugs (bDMARDs), tender joint count (TJC), and age; high HRQOL predictors were PsA Impact of Disease, Beck Depression Inventory (BDI), height, TJC, and BMI (kg/m2). In axSpA, the main LDA predictors were Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), pretreatment with bDMARDs, C-reactive protein, Assessment of SpondyloArthritis international Society Health Index (ASAS HI), and height; high HRQOL predictors were ASAS HI, BDI, BMI, height, and age. XAI provides significant value by enabling explanations of individual patient predictions and their visualizations. This modeling approach may help in the development of a clinical decision support system for patient management.