Accurate and efficient detection of small-scale targets on dynamic water surfaces remains a critical challenge in the deployment of unmanned surface vehicles (USVs) for maritime applications. Complex background interference—such as wave motion, sunlight reflections, and low contrast—often leads to missed or false detections, particularly when using conventional convolutional neural networks. To address these issues, this study introduces LMS-YOLO, a lightweight detection framework built upon the YOLOv8n architecture and optimized for real-time marine object recognition. The proposed network integrates three key components: (1) a C2f-SBS module incorporating StarNet-based Star Blocks, which streamlines multi-scale feature extraction while reducing parameter overhead; (2) a Shared Convolutional Lightweight Detection Head (SCLD), designed to enhance detection precision across scales using a unified convolutional strategy; and (3) a Mixed Local Channel Attention (MLCA) module, which reinforces context-aware representation under complex maritime conditions. Evaluated on the WSODD and FloW-Img datasets, LMS-YOLO achieves a 5.5% improvement in precision and a 2.3% gain in mAP@0.5 compared to YOLOv8n, while reducing parameter count and computational cost by 37.18% and 34.57%, respectively. The model operates at 128 FPS on standard hardware, demonstrating its practical viability for embedded deployment in marine perception systems. These results highlight the potential of LMS-YOLO as a deployable solution for high-speed, high-accuracy marine object detection in real-world environments.

The rapid expansion of Internet of Things (IoT) systems and networks has led to increased challenges regarding security and system reliability. Anomaly detection has become a critical task for identifying system flaws, cyberattacks, and failures in IoT environments. This study proposes a hybrid deep learning (DL) approach combining Autoencoders (AE) and Long Short-Term Memory (LSTM) networks to detect anomalies in real-time within IoT networks. In this model, normal data trends were learned in an unsupervised manner using an AE, while temporal dependencies in time-series data were captured through the use of an LSTM network. Experiments conducted on publicly available IoT datasets, namely the Kaggle IoT Network Traffic Dataset and the Numenta Anomaly Benchmark (NAB) dataset, demonstrate that the proposed hybrid model outperforms conventional machine learning (ML) algorithms, such as Support Vector Machine (SVM) and Random Forest (RF), in terms of accuracy, precision, recall, and F1-score. The hybrid model achieved a recall of 96.2%, a precision of 95.8%, and an accuracy of 97.5%, with negligible false negatives and false positives. Furthermore, the model is capable of handling real-time data with a latency of just 75 milliseconds, making it suitable for large-scale IoT applications. The performance evaluation, which utilized a diverse set of anomaly scenarios, highlighted the robustness and scalability of the proposed model. The Kaggle IoT Network Traffic Dataset, consisting of approximately 630,000 records across six months and 115 features, along with the NAB dataset, which includes around 365,000 sensor readings and 55 features, provided comprehensive data for evaluating the model’s effectiveness in real-world conditions. These findings suggest that the hybrid DL framework offers a robust, scalable, and efficient solution for anomaly detection in IoT networks, contributing to enhanced system security and dependability.

Handwritten digit classification represents a foundational task in computer vision and has been widely adopted in applications ranging from Optical Character Recognition (OCR) to biometric authentication. Despite the availability of large benchmark datasets, the development of models that achieve both high accuracy and computational efficiency remains a central challenge. In this study, the performance of three representative machine learning paradigms—Chi-Squared Automatic Interaction Detection (CHAID), Generative Adversarial Networks (GANs), and Feedforward Deep Neural Networks (FFDNNs)—was systematically evaluated on the Modified National Institute of Standards and Technology (MNIST) dataset. The assessment was conducted with a focus on classification accuracy, computational efficiency, and interpretability. Experimental results demonstrated that deep learning approaches substantially outperformed traditional Decision Tree (DT) methods. GANs and FFDNNs achieved classification accuracies of approximately 97%, indicating strong robustness and generalization capability for handwritten digit recognition tasks. In contrast, CHAID achieved only 29.61% accuracy, highlighting the limited suitability of DT models for high-dimensional image data. It was further observed that, despite the computational demand of adversarial training, GANs required less time per epoch than FFDNNs when executed on modern GPU architectures, thereby underscoring their potential scalability. These findings reinforce the importance of model selection in practical deployment, particularly where accuracy, computational efficiency, and interpretability must be jointly considered. The study contributes to the ongoing discourse on the role of artificial intelligence (AI) in pattern recognition by providing a comparative analysis of classical machine learning and deep learning approaches, thereby offering guidance for the development of reliable and efficient digit recognition systems suitable for real-world applications.

The integration of artificial intelligence (AI) in precision agriculture has facilitated significant advancements in crop health monitoring, particularly in the early identification and classification of foliar diseases. Accurate and timely diagnosis of plant diseases is critical for minimizing crop loss and enhancing agricultural sustainability. In this study, an interpretable deep learning model—referred to as the Multi-Crop Leaf Disease (MCLD) framework—was developed based on a Convolutional Neural Network (CNN) architecture, tailored for the classification of tomato and grapevine leaf diseases. The model architecture was derived from the Visual Geometry Group Network (VGGNet), optimized to improve computational efficiency while maintaining classification accuracy. Leaf image datasets comprising healthy and diseased samples were employed to train and evaluate the model. Performance was assessed using multiple statistical metrics, including classification accuracy, sensitivity, specificity, precision, recall, and F1-score. The proposed MCLD framework achieved a detection accuracy of 98.40% for grapevine leaf diseases and a classification accuracy of 95.71% for tomato leaf conditions. Despite these promising results, further research is required to address limitations such as generalizability across variable environmental conditions and the integration of field-acquired images. The implementation of such interpretable AI-based systems is expected to substantially enhance precision agriculture by supporting rapid and accurate disease management strategies.

Accurate forecasting of Gross Domestic Product (GDP) growth remains essential for supporting strategic economic policy development, particularly in emerging economies such as Indonesia. In this study, a hybrid predictive framework was constructed by integrating fuzzy logic representations with machine learning algorithms to improve the accuracy and interpretability of GDP growth estimation. Annual macroeconomic data from 1970 to 2023 were utilised, and 19 input features were engineered by combining numerical economic indicators with fuzzy-based linguistic variables, along with a forecast label generated via the Non-Stationary Fuzzy Time Series (NSFTS) method. Six supervised learning models were comparatively assessed, including Random Forest (RF), Support Vector Regression (SVR), eXtreme Gradient Boosting (XGBoost), Huber Regressor, Decision Tree (DT), and Multilayer Perceptron (MLP). Model performance was evaluated using Mean Absolute Error (MAE) and accuracy metrics. Among the tested models, the RF algorithm demonstrated superior performance, achieving the lowest MAE and an accuracy of 99.45% in forecasting GDP growth for 2023. Its robustness in capturing non-linear patterns and short-term economic fluctuations was particularly evident when compared to other models. These findings underscore the RF model's capability to serve as a reliable tool for economic forecasting in data-limited and volatile macroeconomic environments. By enabling more precise GDP growth predictions, the proposed hybrid framework offers a valuable decision-support mechanism for policymakers in Indonesia, contributing to more informed resource allocation, proactive economic intervention, and long-term development planning. The methodological innovation of integrating NSFTS with machine learning extends the frontier of data-driven macroeconomic modelling and provides a replicable template for forecasting applications in other emerging markets.

An advanced machine learning (ML)-driven web platform was developed and deployed to automate the identification of rice insect pests, addressing limitations associated with traditional pest detection methods and conventional ML algorithms. Historically, pest identification in rice cultivation has relied on expert evaluation of pest species and their associated crop damage, a process that is labor-intensive, time-consuming, and prone to inaccuracies, particularly in the misclassification of pest species. In this study, a subset of the publicly available IP102 benchmark dataset, consisting of 7,736 images across 12 rice pest categories, was curated for model training and evaluation. Two classification models—a Support Vector Machine (SVM) and a deep Convolutional Neural Network (CNN) based on the Inception_ResNetV2 architecture—were implemented and assessed using standard performance metrics. Experimental results demonstrated that the Inception_ResNetV2 model significantly outperformed SVM, achieving an accuracy of 99.97%, a precision of 99.46%, a recall of 99.81%, and an F1-score of 99.53%. Owing to its superior performance, the Inception_ResNetV2 model was integrated into a web-based application designed for real-time pest identification. The deployed system exhibited an average response time of 5.70 seconds, representing a notable improvement in operational efficiency and usability over previous implementations. The results underscore the potential of artificial intelligence in transforming agricultural practices by enabling accurate, scalable, and timely pest diagnostics, thereby enhancing pest management strategies, mitigating crop losses, and supporting global food security initiatives.

Image segmentation remains a foundational task in computer vision, remote sensing, medical imaging, and object detection, serving as a critical step in delineating object boundaries and extracting meaningful regions from complex visual data. However, conventional segmentation methods often exhibit limited robustness in the presence of noise, intensity inhomogeneity, and intricate region geometries. To address these challenges, a novel segmentation framework was developed, integrating fuzzy logic with geometric principles. Uncertainty and overlapping intensity distributions within regions were modeled through fuzzy membership functions, allowing for more flexible and resilient region characterization. Simultaneously, geometric principles—specifically image gradients and curvature—were incorporated to guide boundary evolution, thereby improving delineation precision. A fuzzy energy functional was constructed to jointly optimize region homogeneity, edge preservation, and boundary smoothness. This functional was minimized through an iterative level-set evolution process, allowing dynamic adaptation to varying image characteristics while maintaining computational efficiency. The proposed model demonstrated robust performance across diverse image modalities, including those with high noise levels and complex regional structures, outperforming traditional methods in terms of segmentation accuracy and stability. Its applicability to tasks demanding high-precision region-based analysis highlights its potential for widespread deployment in advanced imaging applications.

As market saturation and competitive pressure intensify within the banking sector, the mitigation of customer churn has emerged as a critical concern. Given that the cost of acquiring new clients substantially exceeds that of retaining existing ones, the development of highly accurate churn prediction models has become imperative. In this study, a hybrid customer churn prediction model was developed by integrating Sentence Transformers with a stacking ensemble learning architecture. Customer behavioral data containing textual content was transformed into dense vector representations through the use of Sentence Transformers, thereby capturing contextual and semantic nuances. These embeddings were combined with normalized structured features. To enhance predictive performance, a stacking ensemble method was employed to integrate the outputs of multiple base models, including random forest, Gradient Boosting Tree (GBT), and Support Vector Machine (SVM). Experimental evaluation was conducted on real-world banking data, and the proposed model demonstrated superior performance relative to conventional baseline approaches, achieving notable improvements in both accuracy and the area under the curve (AUC). Furthermore, the analysis of model outputs revealed several salient predictors of customer attrition, such as anomalous transaction behavior, prolonged inactivity, and indicators of dissatisfaction with customer service. These insights are expected to inform the development of targeted intervention strategies aimed at strengthening customer retention, improving satisfaction, and fostering long-term institutional growth and stability.

Significant advancements in artificial intelligence (AI) have transformed clinical decision-making, particularly in disease detection and management. Endometriosis, a chronic and often debilitating gynecological disorder, affects a substantial proportion of reproductive-age women and is associated with pelvic pain, infertility, and a reduced quality of life. Despite its high prevalence, non-invasive and accurate diagnostic methods remain limited, frequently resulting in delayed or missed diagnoses. In this study, a novel diagnostic framework was developed by integrating deep learning (DL) with explainable artificial intelligence (XAI) to address existing limitations in the early and non-invasive detection of endometriosis. Abdominopelvic magnetic resonance imaging (MRI) data were obtained from the Crestview Radiology Center in Victoria Island, Lagos State. Preprocessing procedures, including Digital Imaging and Communications in Medicine (DICOM)-to-PNG conversion, image resizing, and intensity normalization, were applied to standardize the imaging data. A U-Net architecture enhanced with a dual attention mechanism was employed for lesion segmentation, while Gradient-weighted Class Activation Mapping (Grad-CAM) was incorporated to visualize and interpret the model’s decision-making process. Ethical considerations, including informed patient consent, fairness in algorithmic decision-making, and mitigation of data bias, were rigorously addressed throughout the model development pipeline. The proposed system demonstrated the potential to improve diagnostic accuracy, reduce diagnostic latency, and enhance clinician trust by offering transparent and interpretable predictions. Furthermore, the integration of XAI is anticipated to promote greater clinical adoption and reliability of AI-assisted diagnostic systems in gynecology. This work contributes to the advancement of non-invasive diagnostic tools and reinforces the role of interpretable DL in the broader context of precision medicine and women's health.