Transforming underlying cardiovascular risk into actionable clinical decisions remains a major challenge in contemporary healthcare. Despite advances in cardiology, early-stage cardiovascular disease often remains undetected, which hinders timely intervention and leads to preventable deaths. To overcome this problem, this study presents an explainable machine learning framework for the early diagnosis of cardiovascular disease (CVD). Initially, this study examined several data-balancing strategies, for example, SMOTE (Synthetic Minority Over-sampling Technique), SMOTETomek (Synthetic Minority Over-sampling Technique + Tomek Links), Tomek Links, ADASYN (Adaptive Synthetic Sampling), and SMOTE-ENN (Synthetic Minority Over-sampling Technique-Edited Nearest Neighbors) within the data-preprocessing pipeline. We proposed a novel Adaptive Inertia Weight Gorilla Troops Optimizer (AIW-GTO) to overcome classical GTO’s (Gorilla Troops Optimizer) unstable convergence by adaptively controlling step sizes. It uses large exploratory steps early for wide search and smaller steps later for fine-tuned local optimization, which ensures stable convergence and enhanced optimization accuracy. Several machine learning techniques, namely XGBoost, Random Forest, SVM (Support Vector Machine), LightGBM (Light Gradient Boosting Machine), and MLP (Multilayer Perceptron) classifier, were evaluated on the multi-regional UCI heart disease dataset. The experimental findings revealed that, by integrating AIW-GTO Optimization and class imbalance mitigation, LightGBM and XGBoost individually achieved a benchmark accuracy of 93.48% and 91.85%, respectively. Moreover, a weighted ensemble of them further improved the accuracy to 94.02%. Sensitivity analysis further evaluated the model’s ability to perform under incomplete clinical test data. To enhance ethical considerations and clinical trust, SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) were utilized to provide model explainability and identify the most influential features affecting prediction outcomes. Analysis indicated that ECG-related (Electrocardiogram) features, including ST_Slope (exercise-induced ST change) and Oldpeak (ST depression magnitude), emerged as key predictors of CVD risk. Overall, the proposed framework provides a clinically reliable and interpretable approach for early cardiovascular risk assessment to enable proactive patient management.

Manual inspection of onba earing casting defects is not realistic and unreliable, particularly in the case of some micro-level anomalies which lead to major defects on a large scale. To address these challenges, we propose BearFusionNet, an attention-based deep learning architecture with multi-stream, which merges both DenseNet201 and MobileNetV2 for feature extraction with a classification head inspired by VGG19. This hybrid design, figuratively beaming from one layer to another, extracts the enormity of representations on different scales, backed by a pre-preprocessing pipeline that brings defect saliency to the fore through contrast adjustment, denoising, and edge detection. The use of multi-head self-attention enhances feature fusion, enabling the model to capture both large and small spatial features. BearFusionNet achieves an accuracy of 99.66% and Cohen’s kappa score of 0.9929 in Kaggle’s Real-life Industrial Casting Defects dataset. Both McNemar’s and Wilcoxon signed-rank statistical tests, as well as five-fold cross-validation, are employed to assess the robustness of our proposed model. To interpret the model, we adopt Grad-Cam visualizations, which are the state of the art standard. Furthermore, we deploy BearFusionNet as a web-based system for near real-time inference (5–6 s per prediction), which enables the quickest yet accurate detection with visual explanations. Overall, BearFusionNet is an interpretable, accurate, and deployable solution that can automatically detect casting defects, leading to significant advances in the innovative industrial environment.

The rapid growth in available network bandwidth has directly contributed to an exponential increase in mobile data traffic, creating significant challenges for network energy consumption. Also, with the extraordinary growth of mobile communications, the data traffic has dramatically expanded, which has led to massive grid power consumption and incurred high operating expenditure (OPEX). However, the majority of current network designs struggle to efficiently manage a massive amount of data using little power, which degrades energy efficiency performance. Thereby, it is necessary to have an efficient mechanism to reduce power consumption when processing large amounts of data in network data centers. Utilizing renewable energy sources to power the Cloud Radio Access Network (C-RAN) greatly reduces the need to purchase energy from the utility grid. In this paper, we propose a bandwidth-aware hybrid energy-powered C-RAN that focuses on throughput and energy efficiency (EE) by lowering grid usage, aiming to enhance the EE. This paper examines the energy efficiency, spectral efficiency (SE), and average on-grid energy consumption, dealing with the major challenges of the temporal and spatial nature of traffic and renewable energy generation across various network setups. To assess the effectiveness of the suggested network by changing the transmission bandwidth, a comprehensive simulation has been conducted. The numerical findings support the efficacy of the suggested approach.

Urban parking problems worsen traffic jams, gas use, and pollution. Old parking systems often lack up-to-date space information, which annoys drivers and wastes their time. This research presents a smart IoT-enabled real-time space monitoring and booking system applicable to both urban parking management and Smart Manufacturing logistics environments, including loading bay coordination and Automated Guided Vehicle (AGV) docking station management. The system employs ultrasonic and IR sensors, managed by an Arduino UNO, to identify vehicles and track space availability. A servo-motor regulates entry. Slot data is presented on a Liquid Crystal Display screen and accessible through a mobile app. The tests suggest that the system is accurate (98.67%) and reduces entry and exit times to 1–2 s for gate actuation, and it can handle increased demand well. Proteus simulations support the system’s reliability. Real-time updates to slot availability improve the user experience and overall system efficiency in both urban and industrial deployment scenarios.

Traditional Intrusion Detection Systems (IDSs) that rely on fixed signatures or basic machine learning often struggle with sophisticated, multi-stage cyberattacks and previously unknown threats. To fix these problems, this paper introduces IntrusionNet, a mixed deep learning system that combines Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Autoencoders in a two-part design. Differing from typical stacked models, IntrusionNet works on two levels at the same time. First, a supervised CNN-RNN process pulls spatial-temporal data from traffic flows to sort well-known attack patterns. Second, an unsupervised Autoencoder process spots new anomalies by looking at reconstruction error limits. This approach allows the automatic learning of threat traits as they change, without needing someone to do it by hand. The system was tested on the UNSW-NB15 data set, picked because it realistically includes many kinds of attacks, like Fuzzers, Shellcode, and Worms. Tests show that IntrusionNet gets an accuracy of 98.80% and an F1-score of 0.985, doing better than other systems, especially with less common attack types. Also, tests using Precision-Recall (PR) analysis and False Positive Rate (FPR) measurements prove that the model handles class imbalance well, which is key for real-world security. The suggested system can be scaled up easily and performs calculations fast, making it a possible key part of real-time detection in Security Information and Event Management (SIEM) systems.

The context of recognizing handwritten city names, this research addresses the challenges posed by the manual inscription of Bangladeshi city names in the Bangla script. In today’s technology-driven era, where precise tools for reading handwritten text are essential, this study focuses on leveraging deep learning to understand the intricacies of Bangla handwriting. The existing dearth of dedicated datasets has impeded the progress of Bangla handwritten city name recognition systems, particularly in critical areas such as postal automation and document processing. Notably, no prior research has specifically targeted the unique needs of Bangla handwritten city name recognition. To bridge this gap, the study collects real-world images from diverse sources to construct a comprehensive dataset for Bangla Hand Written City name recognition. The emphasis on practical data for system training enhances accuracy. The research further conducts a comparative analysis, pitting state-of-the-art (SOTA) deep learning models, including EfficientNetB0, VGG16, ResNet50, DenseNet201, InceptionV3, and Xception, against a custom Convolutional Neural Networks (CNN) model named “Our CNN.” The results showcase the superior performance of “Our CNN,” with a test accuracy of 99.97% and an outstanding F1 score of 99.95%. These metrics underscore its potential for automating city name recognition, particularly in postal services. The study concludes by highlighting the significance of meticulous dataset curation and the promising outlook for custom CNN architectures. It encourages future research avenues, including dataset expansion, algorithm refinement, exploration of recurrent neural networks and attention mechanisms, real-world deployment of models, and extension to other regional languages and scripts. These recommendations offer exciting possibilities for advancing the field of handwritten recognition technology and hold practical implications for enhancing global postal services.