Nighttime high-power aircraft engine run-ups represent a critical operational activity with potential implications for airport management and environmental assessment; yet, their quantitative prediction remains limited in existing literature. The objective of this study is to develop and evaluate a data-driven machine learning framework capable of accurately predicting nighttime high-power engine run-up characteristics using historical operational data. A supervised learning methodology is adopted using an open-access dataset comprising approximately 860 nighttime run-up records. The proposed framework includes data preprocessing, temporal feature engineering (month, season, and quarter), exploratory data analysis, correlation assessment, and comparative model development using Linear Regression, Random Forest, Gradient Boosting, Extreme Gradient Boosting (XGBoost), and an Artificial Neural Network (multilayer perceptron). Model performance is evaluated using coefficient of determination (R²), root mean square error (RMSE), mean absolute error, and mean absolute percentage error, supported by predicted-versus-actual comparisons and learning-curve analysis. The results demonstrate that ensemble-based models outperform linear and neural network approaches, with XGBoost achieving the highest predictive accuracy on the testing dataset (R² ≈ 0.90, RMSE ≈ 2.15). Learning curves further indicate stable validation performance with increasing training data size. The proposed machine learning framework provides a structured and data-driven approach for predicting nighttime high-power aircraft engine run-up activity within the analyzed operational dataset, offering a valuable analytical tool for data-driven assessment of airport ground operations.

The rising demand for poultry products makes it necessary to implement artificial egg incubation systems, although backyard poultry and developing countries are hindered by manual incubation or expensive commercial incubation facilities. To overcome this, this paper introduces an artificial intelligence-optimized egg incubation system with energy awareness in poultry farming, offering a low-cost, intelligent, and optimal incubation solution. The proposed system is designed using an Arduino-embedded system with temperature and humidity sensors, heating, and fans, with a GSM notification system for real-time monitoring of incubation conditions. The proposed system differs from traditional incubation systems, which are based on threshold values, as this system uses a machine learning approach-based predictive control system, which predicts changes in environmental factors to control actuators proactively. Based on anticipated changes in temperatures and humidity, this approach avoids unnecessary activation of heating and fans, which cuts down on energy consumption with optimal incubation conditions. Experimental outcome shows better temperature regulation, improved incubation success, and significant energy savings over traditional ON/OFF control systems. The outcome validates that artificial intelligence is beneficial in developing intelligent, adaptive, and energy-conscious egg incubation systems for backyard poultry farming.

Artificial intelligence (AI) has the potential to transform diabetic retinopathy (DR) screening, yet its feasibility and equity in low- and middle-income countries (LMICs), particularly sub-Saharan Africa (SSA), remain insufficiently examined. We conducted a systematic review and meta-analysis following PRISMA 2020 guidelines, searching PubMed and Google Scholar for peer-reviewed studies on AI-based DR screening in LMICs, with emphasis on SSA. Data on diagnostic performance, dataset composition, feasibility, and equity were extracted, and sensitivity and specificity were pooled using random-effects meta-analysis; qualitative synthesis identified infrastructural, economic, and ethical factors shaping implementation. Fifty studies met the inclusion criteria. Pooled specificity and sensitivity were 87.63% and 91.91%, respectively, both with very high heterogeneity (I² = 99.6% and 98.1%, respectively; P < 0.001), indicating that performance differences are unlikely to be due to chance alone. In LMIC validation settings, pooled specificity was higher (89.15%, 95% confidence interval (CI): 85.31-92.99; I² = 84.98%) but sensitivity was lower (91.23%, 95% CI: 88.14-94.31; I² = 96.4%) compared with high income countries (HIC) settings (specificity 82.80%, 95% CI: 75.35-90.25; I² = 99.81%; sensitivity 94.91%, 95% CI: 91.26-98.56; I² = 88.6%). This pattern suggests that models validated in LMICs may be more conservative in ruling out DR, whereas models validated in HICs may be more adept at detecting true positives. Key barriers to AI implementation in SSA included inadequate digital infrastructure, workforce shortages, high proprietary system costs, and underrepresentation of SSA researchers in leadership roles. Overcoming these challenges will require coordinated investment in local capacity, digital infrastructure, equitable research partnerships, and the adoption of open, transparent, and inclusive AI development, underpinned by cross-border governance. With such measures, AI could become a sustainable and equitable tool for reducing preventable vision loss across SSA and other resource-constrained regions.

Multichannel surface electromyography-based hand gesture recognition plays an important role in the development of intuitive prosthetic and human-machine interfaces. It requires high accuracy, minimal noise degradation, and electrode drift tolerance for real-world deployment. The goal is to find the best tuned classifier configurations that balance accuracy and real-world robustness, comparing grid search, random search, and Bayesian optimization for the four classical classifiers: K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM). Their performance is evaluated by repeated stratified 5 × 5 cross-validation with 25 runs and leave-one-subject-out testing. Furthermore, we use a multi-objective Pareto front analysis considering three factors: (1) accuracy, (2) adaptation to simulated drift, and (3) relative accuracy degradation under noise. KNN reaches stable 98.4% accuracy on all optimizers, but Bayesian optimization performs better than other optimization techniques in terms of accuracy and other performance metrics when using SVM, DT, and RF classifiers. Additionally, Bayesian optimization gets KNN into the Pareto front with 98.44% accuracy, 98.9% adaptability, and a minimal 0.10% decrease in noise, which likewise clearly improves SVM and RF. These results provide guidance about how to choose effective, real-time electromyography classifiers not only for accuracy but also through trade-offs of accuracy, adaptability, and noise tolerance in prosthetic and rehabilitation systems.

The evaluation of credit risk remains an important task in the financial sector, requiring high precision and interpretability in order to neutralize default risks. This work presents a comprehensive and hybrid classification model to forecast credit risk on the German credit dataset. The model proposed is a combination of traditional machine learning classifiers, i.e., the Decision Tree (C4.5) and Support Vector Machine with an evolutionary Hybrid Genetic Algorithm as a feature selector, thus enhancing classification performance. Besides single models, we utilize ensemble learning approaches, i.e., bagging, with and without feature selection, to investigate performance on varying iterations of bagging. Experiments are performed on both quantitative and qualitative forms of the dataset, split into five train-test ratios (60:40, 70:30, 75:25, 80:20, 85:15), in order to examine the stability and generalization ability of the models. The Hybrid Genetic Algorithm-based proposed classifiers minimize misclassification errors more effectively, especially when the training set is small, while, concurrently, obtaining low Type I and Type II error rates. Ensemble models with feature selection (Feature Selection-Hybrid Bagging) also outperformed the performance of conventional bagging models in all partitions of the data consistently. The evaluation metrics like Accuracy, Precision, Recall, F1-score, Area Under the Receiver Operating Characteristic Curve, and Log Loss confirm the efficacy of this hybrid approach. The results confirm that the combination of genetic algorithms for feature optimization and conventional classifiers leads to a strong and interpretable solution to credit risk classification problems. The hybrid model is a reliable decision-support system for financial institutions aiming to automate and improve credit scoring procedures.

The rapid expansion of Internet of Things (IoT) networks has heightened the risk of sophisticated malware attacks due to the inherent vulnerabilities of resource-constrained devices. This study proposes the SEQIRE model - a novel compartmental framework inspired by epidemiological modeling - to analyze, detect, and control malware propagation in IoT ecosystems. The model extends classical SEIR-type structures by incorporating two additional compartments: Quarantined (Q) and Enhanced Monitoring (E*), enabling a more accurate representation of real-world cybersecurity interventions. A system of nonlinear ordinary differential equations was formulated to describe device transitions across six compartments: Susceptible, Exposed, Quarantined, Infected, Recovered, and Enhanced Monitoring. Theoretical analyses, including boundedness, positivity, and Lyapunov-based global stability, were conducted to ensure model feasibility. The basic reproduction number R0​ was derived using the Next-Generation Matrix method as a threshold condition for malware persistence. Numerical simulations using Python’s Runge-Kutta (RK45) method over a 200-day window validated the model's predictive performance. Key parameters - quarantine rate (γ), recovery rate (μ), and monitoring activation rate (θ) - were varied to assess their influence on infection suppression. Results indicate that increasing these parameters significantly reduces infection peaks and accelerates system stabilization. The SEQIRE model offers both analytical rigor and practical utility for dynamic threat mitigation in IoT networks. Its flexibility allows for deployment in real-time detection systems and supports decision-making for resource-optimized cybersecurity strategies.

Currently, artificial intelligence represents a disruptive technology with an extensive range of applications that has become popular due to the efficiency and speed with which it performs analysis, obtains results, and optimizes processes of diverse nature. As an ally of humans, it has proven to be exceptional. However, its application also generates controversy due to the nature of its use and development, as well as the type and origin of the data used to train it. This has raised risk scenarios for the dignity, integrity, and protection of the fundamental rights of individuals that must be regulated through a solid legal framework that considers aspects of governance, security, and ethical implications derived from its implementation. This paper addresses the origins and concept of artificial intelligence, some of its applications and general benefits, some of the potential risks derived from its application, as well as an updated analysis regarding the status of some of the main initiatives that have emerged in the international arena involved in its regulation with the aim of identifying common approaches, guidelines, and principles that are being considered by governments and that can establish guidelines towards a harmonized regulation of AI for the countries that cannot be postponed.

Number plate recognition is an important component of smart transportation systems as it plays a significant role in automated traffic management, law enforcement, and road safety. In order to achieve accurate and effective recognition under various types of settings, this paper provides a thorough analysis of number plate preprocessing and recognition algorithms. Our system leverages computer vision and object detection techniques to accurately identify vehicles and extract their license plates. To ensure precise number plate recognition, we employ methods such as skew correction, noise reduction, grayscale conversion, and optical character recognition. Common issues with number plate recognition like slanted orientations and varying lighting conditions are addressed. Through the integration of these techniques, our system achieves a recognition accuracy of 98.1%, outperforming several existing approaches. Our approach attempts to offer insights into enhancing the accuracy and robustness of number plate recognition systems by assessing current approaches and their effectiveness.

Coffee prices are known for their unpredictability, shaped by different aspects such as weather, trade dynamics, and seasonal demand. In this study, several forecasting methods were applied to monthly coffee price data to find a model that balances accuracy with the ability to handle volatility. Techniques ranged from classic time series models to deep learning and hybrid approaches. The best performer was the long short-term memory model, which effectively captured complex, nonlinear patterns and volatility in coffee price movements, resulting in the lowest error rates among all models tested. While traditional statistical models captured trends, they often struggled to keep up with price variations. The long short-term memory model was trained for a maximum of 200 epochs; however, training effectively converged and was stopped early around the 55th epoch, as the loss stabilized at a low value. This early convergence further shows the model’s efficiency in learning complex, nonlinear patterns from the coffee price data. This study would help the policy-makers, agriculturer, and traders.