Band 34, Heft 1

With the acceleration of industrialization and the increase in the complexity of equipment, it is particularly important to accurately predict and effectively maintain equipment failures. In response to the problem of difficulty in accurately achieving predictive maintenance and the severely limited application range of new real-time disassembly sensors, this article conducts research on the synergistic effect of the new real-time disassembly sensors and artificial intelligence technology. First, relevant parameter data of a certain enterprise’s generator can be collected on-site, and the Pearson correlation coefficient method can be used to calculate the correlation between the generator data parameters and the target fault type, ensuring the degree of correlation in extracting data features. Then, based on the gated recurrent unit (GRU) model, the article applied the particle swarm optimization (PSO) algorithm to optimize the parameters of the GRU network and used the support vector machine (SVM) model to optimize the classification function of the network output. Finally, the optimized GRU model can be applied to predict the fault types of generators, and based on this, it can be applied to the energy industry, agriculture, medical and health, and transportation industries to verify the application scalability of artificial intelligence and new real-time disassembly sensors. The experimental results show that the optimized GRU model combined with the new real-time disassembly sensor achieved an average accuracy of 0.94 in generator fault prediction, with a cost loss rate of only 3.03%, a decrease of 5.91% compared to the single new real-time disassembly sensor. The combination of artificial intelligence technology for precise predictive maintenance of generators greatly reduces maintenance costs and overcomes the limitations of individual real-time sensor disassembly. This has to some extent expanded the application scope and promoted the intelligent development of various industries.

Compared with traditional agriculture, greenhouse planting can more accurately control the growth environment, thereby improving the yield and agricultural product quality. However, traditional greenhouse environments (GhEs) present a number of challenges, including inflexibility in monitoring and wiring, difficulty in management, and high labor costs. To improve the limitations of traditional GhEs and enhance the accuracy of GhE monitoring and control systems, a sensor-based GhE monitoring and control system is designed. In addition, a prediction model for GhE monitoring is constructed using a backpropagation neural network to better predict nonlinear factors such as humidity, temperature, and light intensity in the GhE. Simultaneously, an improved fuzzy proportional integral derivative (PID) controller is utilized to address issues such as fuzziness and uncertainty in GhEs. The results show that the temperature error of the greenhouse environment monitoring and control system based on improved fuzzy PID and neural network algorithms is 0.35–5.04%, the humidity error is −1.3 to 1.65%, and the lighting error is −3.5 to −0.79%. Comprehensive data show that the greenhouse environmental monitoring and control (GEMC) system based on improved fuzzy PID and neural network algorithms effectively improves the accuracy of environmental monitoring and control. The GEMC system, which is based on improved fuzzy PID and neural network algorithms, has facilitated the advancement of agricultural technology in China. It has also provided support and a reference point for GhE monitoring and agricultural production.

The Egyptian cobra is among the deadliest snake species, capable of causing death within a short span of 15 min. Also, every snake species has its own anti-venom type. So, a quick identifying the Egyptian Cobra bite from other snake species is a challenging and critical task. This research employs Internet of things (IoT) and deep learning methods to precisely recognize bites of Egyptian cobra, in the real-time, by analyzing images of the bite marks. We deploy IoT-enabled wearable devices equipped with sensors capable of detecting snake bites, whereas these sensors measure changes in physiological parameters indicative of a snakebite, such as heart rate, blood pressure, and temperature sensors based on our proposed mathematical algorithm. Also, we present a real case study in which we used our mathematical algorithm to determine based on its sensor readings whether the victim was exposed to a snake bite or not in the real-time. These wearable devices can be worn by individuals working or living in areas prone to snake encounters, such as farmers. When a snake bite occurs, the IoT sensors embedded in the wearable devices will immediately detect the bite and transmit real-time data, including vital information about the bite marks, to a central monitoring system or victim relative. Also, we assembled a dataset consisting of 500 images depicting Egyptian cobra bites and 600 images of bites from various other snake species indigenous to Egypt. To bolster the model’s trustworthiness and facilitate understanding of its decisions, we employed the contemporary method of explainable deep learning. Also, notably, our methodology yielded an accuracy of 90.9%.

Identification of cyber threats is crucial and significant for determining substantial security techniques and approaches. This research illustrates a brief discussion of cyberspace challenges and threats in a disruptive era alongside comprehensive approaches in mitigating the risk of cyber threats. Additionally, the aim of this research is to provide beneficial approaches on how to handle cyber threats in detail. For example, threats and attacks may be caused in the absence of legislation, ethical standardization, support system, and lack of access control. The governance system, therefore, will put a lot of effort into communicating, identifying, and enforcing the principles of security to moderate risk. The Metaheuristic algorithms are stimulated by the human brain, so implementing Artificial Intelligence (AI) that assists the Neural Network to mimic the behaviour of the human brain is important to predict significant outcomes. In this study, the author investigates and analyses the rapid growth of cyber threats to outline the solutions. The aim of this study is to contribute to cyber security optimization through implementing AI methodologies.

Diabetic retinopathy (DR) is a leading cause of blindness worldwide, necessitating early detection to prevent severe visual impairment. Despite numerous proposed classification techniques, challenges persist due to the high parameter count of deep learning algorithms, imbalanced datasets, and limited performance. This study introduces a novel framework for DR classification that leverages multi-view deep features, multilinear whitened principal component analysis, tensor exponential discriminant analysis, synthetic minority oversampling technique, and deep random forest. We evaluated this architecture using the APTOS blindness dataset under a standard protocol. The results demonstrate that our architecture significantly improves classification accuracy, surpassing existing methods. Our contributions highlight a promising approach for enhancing DR classification performance.

Multi-objective scheduling problems are inherently complex due to the need to balance competing objectives, such as minimizing the total weighted completion time, reducing the number of delayed jobs, and minimizing the maximum weighted delay. To address these challenges, this article introduces the meerkat clan algorithm (MCA), inspired by the dynamic, cooperative, and adaptive behaviors of meerkats, which enhances the exploration and exploitation of solution spaces. The MCA is further integrated with the traditional branch-and-bound (BAB) method, utilizing it as an upper bound to significantly improve the accuracy and efficiency of the solutions. Comprehensive computational experiments were conducted to evaluate the MCA’s performance against state-of-the-art algorithms, including the bald eagle search optimization algorithm (BESOA) and the standalone BAB method. The MCA demonstrated superior scalability and efficiency, effectively solving problems involving up to n = 30,000 jobs, whereas the BESOA was limited to handling instances with n = 1,000 jobs. Additionally, the integration of MCA with the BAB method achieved exceptional precision and efficiency for smaller problem instances, handling up to n = 13 jobs effectively. The results underscore the MCA algorithm’s potential as a robust solution for multi-objective scheduling problems, combining speed and accuracy to outperform traditional methods. Moreover, the hybrid approach of integrating MCA with BAB provides a flexible and versatile framework capable of addressing a wide range of scheduling scenarios, from small-scale to large-scale applications. These findings position the MCA as a transformative tool for solving complex scheduling problems in both theoretical and practical domains.

A healthy lifestyle encompasses physical, mental, and emotional well-being, with healthcare and nutrition as central components. For those with chronic diseases such as diabetes, effective self-management involves continuous monitoring and dietary adjustments. Understanding the glycemic index (GI) is vital, as it indicates how carbohydrates affect blood sugar levels. Advancements in artificial intelligence have enhanced diabetes management through food image recognition systems, which identify food items and provide nutritional information, helping individuals track their dietary intake and GI consumption effectively. Despite their high performance, existing systems often lack inclusivity for diverse cuisines, such as Moroccan cuisine, which is known for its unique dishes of spices and health benefits. This study addresses these gaps by proposing the first comprehensive Moroccan food dataset, comprising 8,300 images across 70 food categories. The research subsequently proposes an advanced model to enhance food image recognition accuracy using convolutional neural network and attention mechanisms achieving more than 90% accuracy. In addition, estimating the GI values of Moroccan foods will help to raise public awareness of their health implications and facilitate decision-making on dietary self-management. The results demonstrate state-of-the-art performance, indicating promising potential for the first GI estimation of Moroccan food images.

Due to robustness and efficiency, naive Bayes (NB) remains among the top ten data mining algorithms. However, the required conditional independence assumption more or less limits its classification performance. Of numerous approaches to improving NB, structure extension and instance weighting have both achieved remarkable improvements. To make full use of their complementary and consensus advantages, this article proposes a hybrid modeling approach to combining structure extension with instance weighting. We call the resulting model instance weighted averaged one-dependence estimators (IWAODE). In IWAODE, the dependencies among attributes are modeled by an ensemble of one-dependence estimators, and the corresponding probabilities are estimated from attribute value frequency-weighted training instances. The classification performance of IWAODE is experimentally validated on a large number of datasets.

With the increasingly serious global environmental and energy issues, more countries are using environmental and economic dispatch (EED) models to optimize power systems. To better optimize the power system, an adaptive artificial bee colony (ABC) algorithm based on memory feedback mechanism was put forward to address the EED model, and the adaptive algorithm was used to adaptively adjust the population size. The study also used a benchmark function to set an appropriate population size. In addition, the study also considered both fuel cost and environmental factors in the model, and simultaneously considered four constraint conditions. To evidence the function of the adaptive algorithm, different algorithms were compared in the study. The outcomes denoted that the minimum values of the optimal solution under the Sphere function, Matyas function, and Dixon Price function were 1 × 10 −273 , 1 × 10 −162 , and 1 × 10 −16 , respectively, and their corresponding population sizes were 7, 18, and 20. Under the Sphere function, the minimum average fitness values of the algorithm designed by the research, the ABC algorithm, and the current optimal ABC algorithm were 10 −15 , 10 −4 , and 10 −11 , respectively. Moreover, the algorithm designed by the research tended to flatten out after nearly 30 iterations. The total cost of the adaptive algorithm, ABC algorithm, and the optimal algorithm was 102126.0573 yuan, 113001.0383 yuan, and 109594.9634 yuan, respectively. The pollutant emissions of the three algorithms were 1246.1048 yuan, 1250.5744 yuan, and 1344.3922 yuan, respectively. The adaptive algorithm based on memory feedback mechanism had obvious advantages in solving EED models. The adaptive algorithm proposed by the research achieved adaptive adjustment of population size, improved the operational efficiency of the algorithm, and had certain reference significance for solving other problems.

Vietnam’s stock market is characterized as a frontier market and focuses on emerging market status by 2025. Tisco Advisory’s report showed that Vietnam’s stock market is expecting to draw 4 billion in foreign capital in 2024. Despite the appealing nature of the stock market, forecasting stock prices remains a complex endeavor owing to its fast-paced and fluctuating volatility. Effectively forecasting the fluctuation of stock prices has the potential to mitigate the risk associated with stock investments and enhance the overall investment yield. In this research, we combine the advantages of XGBoost for feature selection with the autoregressive integrated moving average (ARIMAX) time series model for forecasting to improve the accuracy of predicting next-day stock prices. A dual important features selection approach is proposed to extract key features for the ARIMAX model from a pool of 87 technical indicators. To demonstrate the effectiveness of this method, we compared it with four other methods – long-short term memory, genetics algorithms with long-short term memory, XGBoost, and Meta Prophet – in predicting the next day’s closing price of the Vietnam stock index from January 2013 to April 2023. The results indicate that the performance of our method is better than others and suitable for traders to make stock investment decisions.

Smart education environments combine technologies such as big data, cloud computing, and artificial intelligence to optimize and personalize the teaching and learning process, thereby improving the efficiency and quality of education. This article proposes a dual-stream-coded image sentiment analysis method based on both facial expressions and background actions to monitor and analyze learners’ behaviors in real time. By integrating human facial expressions and scene backgrounds, the method can effectively address the occlusion problem in uncontrolled environments. To enhance the accuracy and efficiency of emotion recognition, a multi-task convolutional network is employed for face extraction, while 3D convolutional neural networks optimize the extraction process of facial features. Additionally, the adaptive learning screen adjustment system proposed in this article dynamically adjusts the presentation of learning content to optimize the learning environment and enhance learning efficiency by monitoring learners’ expressions and reactions in real time. By analyzing the experimental results on the Emotic dataset, the emotion recognition model in this article shows high accuracy, especially in the recognition of specific emotion categories. This research significantly contributes to the field of smart education environments by providing an effective solution for real-time emotion recognition.

To foster the advancement of telecommunications enterprise supply chains and facilitate their transition toward global market competitiveness, the author advocates for a novel performance evaluation framework tailored for the integrated circuit industry supply chain. This framework integrates the fuzzy analytic hierarchy process and fuzzy neural network methodologies to devise a comprehensive supplier performance assessment model. Leveraging extensive historical supplier data, the author employs MATLAB’s neural network toolbox for model training and simulation. The results indicate that the error value output after running the validation sample is relatively small. This indicates that the model can be effectively applied to the performance evaluation of common integrated circuit product suppliers in CM company. According to the performance results of the model application, all participating suppliers have performance scores greater than 0.7, indicating that in the supply and service process of butterfly integrated circuit products, the performance evaluation scores of each supplier meet the requirements of qualified suppliers, among them, suppliers DS1, DS8, and DS9 with a comprehensive performance score greater than 0.8 are relatively excellent suppliers. The model’s effectiveness and accuracy have been confirmed, demonstrating its practical applicability to CM Company. Moreover, its insights provide valuable guidance for establishing supplier performance evaluation systems across various product categories within telecommunications enterprises.

In the era of artificial intelligence, the technology of optical character recognition under complex backgrounds has become particularly important. This article investigated how machine learning algorithms can improve the accuracy of text recognition in complex scenarios. By analyzing algorithms such as scale-invariant feature transform, K -means clustering, and support vector machine, a system was constructed to address the challenges of text recognition under complex backgrounds. Experimental results show that the proposed algorithm achieves 7.66% higher accuracy than traditional algorithms, and the built system is fast, powerful, and highly satisfactory to users, with a 13.6% difference in results between the two groups using different methods. This indicates that the method proposed in this study can effectively meet the needs of complex text recognition, significantly improving recognition efficiency and user satisfaction.

Comorbidity, the simultaneous existence of multiple medical conditions in a patient, is a major challenge in healthcare. Comorbidity is highly threatening for healthcare systems, which requires innovative solutions over traditional methods. The medical field is challenged by accurately diagnosing these intertwined diseases of coexisting ailments and anticipating their rise. The current diagnostic approaches are time-consuming and inaccurate, hinder effective treatment, and delay accurate results for the patient. Artificial intelligence can provide an effective method for early prediction of comorbidity risks. In this study, various artificial intelligence models are used, and a clinical dataset of 271 patients is utilized to diagnose comorbidity. In which a hybrid diagnosis model is proposed based on the intersection between machine learning (ML) and feature selection techniques for the detection of comorbidity. Fuzzy decision by opinion score method is utilized as a sophisticated tool to select the most representative ML for prediction. Extensive simulation results showed an accuracy rate of 91.463 using AdaBoost ML. Furthermore, utilizing the fuzzy decision by opinion score technique, we were able to confirm that the best model using all features as well as the chi square and KBest features is the AdaBoost, which scored the smallest value of 0.204 and hence confirm that it is the best selected ML model for comorbidity.

The problem of segmenting retinal blood vessels in fundus images arises from the challenges of accurately detecting and delineating vessels due to their complex structures, varying sizes, and overlapping features. Manual segmentation is time-consuming and prone to human error, leading to inconsistent results. Additionally, existing automated methods often struggle with low-quality images or variations in illumination, hindering their effectiveness. Therefore, there is a pressing need for an efficient and accurate automated system to improve segmentation outcomes for better diagnosis of retinal diseases. This study proposes a fully automated model for blood vessel segmentation in retinal fundus images, addressing key challenges such as poor image quality, weak vessel detection, and inhomogeneity in contrast. Macular degeneration and diabetic retinopathy are major causes of vision impairment, making accurate retinal analysis crucial. The proposed model enhances image quality through a novel pre-processing pipeline that includes logarithmic contrast enhancement, noise reduction using an improved complex wavelet transform with shrinkage, and anisotropic diffusion filtering for edge enhancement. The segmentation method combines morphological operations with an optimized Canny edge detector, effectively identifying and segmenting blood vessels. This approach aims to improve the accuracy and efficiency of retinal image analysis, overcoming the limitations of manual segmentation and complex vascular structures. The results obtained from the DRIVE dataset achieved high values for accuracy (Acc, 99%), sensitivity (Sen, 95.83%), specificity (Spe, 98.62%), positive predictive value (PPV, 91.34%), and negative predictive value (NPV, 94%). In addition, the results obtained using the high-resolution fundus dataset were equally satisfactory, achieving an Acc., Sen., Spe., PPV, and NPV of 99.11, 97.97, 98.97, 97.98, and 100%, respectively. These results outperform the gold standard and state-of-the-art schemes to date. The proposed approach increases the performance and reliability of the process of vessel detection in fundus images.

This article discusses the application of computer wireless network database technology in information management, aiming to solve the problem of low efficiency in information management by improving query efficiency and upload rate. Through standardized database technical operations, such as unified addition, deletion, query, and introduction, orderly adjustment and modification of internal big data are carried out to ensure that information transmission can be accurately reflected in the network structure to improve the quality of the system operation. This article uses a fuzzy clustering algorithm to evaluate the information management system comprehensively. It deeply analyzes the significant effects of database technology in improving user experience and enhancing system adaptability through experiments. The experimental results show that the new system is superior to the traditional information management system in data collection, upload efficiency, task management, query statistics, data monitoring, buffering effect, data encryption, access control, resource utilization, bandwidth utilization, data storage efficiency, scalability, reliability, and disaster recovery capability, user experience, and system adaptability. Specifically, the average response time of the new system is less than 2.5 s, the data upload rate reaches 146 Mbps, the user login success rate and the authentication accuracy rate are close to 100%, the resource utilization and network bandwidth utilization reach high levels in different periods, and the data storage efficiency is improved by 0.21. The system’s scalability is improved by 0.29 with the support of database technology. These data show that computer wireless network database technology can significantly improve the efficiency and security of information management and effectively support the needs of high data volume and high-speed industries, such as finance, medical care, and e-commerce.

Predictive maintenance (PdM) is a proactive approach aimed at anticipating the future point of failure for a machine or a component, with the goal of reducing both the frequency and the expenses associated with unplanned downtime. Recent advances in machine learning (ML) techniques have enabled PdM to be more efficient with diverse and successful applications in various manufacturing industries. The support vector machine (SVM), a fundamental ML algorithm, is renowned for its effectiveness in addressing classification and regression tasks. Nevertheless, the successful application of SVM hinges on the careful tuning of its parameters, a process that significantly influences its predictive performance. This research seeks to optimize the selection of the regularization parameter C C and the kernel parameter σ \sigma using metaheuristic methods. It suggests combining the altruistic dragonfly algorithm (ADA) with SVM to enhance the prediction of maintenance failures. The primary motive for integrating altruism into this research is the unprecedented utilization of altruistic principles within this specific area. In addition, ADA-SVM provides a balance between exploration and exploitation. This balance is achieved through the altruistic behavior of dragonflies, where they help each other find better solutions. Therefore, this model is not trapped in the local optimum. The effectiveness of the model ADA-SVM is assessed on aircraft engine sensor data in comparison with other metaheuristic optimization algorithms, namely, genetic algorithms (GA), particle swarm optimization (PSO), grey wolf optimization (GWO) and dragonfly algorithm (DA). The performance of the SVM has been improved significantly by using parameter optimization. Besides, while GA-SVM, PSO-SVM, and DA-SVM models were able to predict engine failures with 95% accuracy, and the GWO-SVM, which demonstrated a good performance in terms of accuracy compared to other metaheuristics algorithms, achieves an accuracy of 97%, the ADA-SVM reached the best accuracy value which is 98%. The findings, thus, reveal that the proposed model outperforms the other models in optimizing SVM parameters, and, therefore, improves the performance of the engines failures prediction.

Tire pressure monitoring systems (TPMS) are essential for vehicle safety and performance as they help detect low tire pressure that impacts fuel efficiency, ride comfort, and overall safety. This study introduces a novel stacking ensemble model to improve the monitoring of nitrogen-filled pneumatic tires. Vibration signals, captured under four conditions such as idle, highspeed, normal, and puncture, using low-cost MEMS accelerometers, are processed to derive autoregressive moving average (ARMA), histogram, and statistical features. The J48 decision tree is employed for feature selection, enhancing classifier accuracy. Experiments with various machine learning classifiers show that the stacking ensemble approach significantly improves classification performance for ARMA (93.75%) and histogram (85.42%) features, thereby achieving higher accuracy than individual classifiers. These findings demonstrate that stacking ensembles can enhance TPMS capabilities, offering a cost-effective and accurate solution for real-time tire pressure monitoring. This advancement contributes to automotive safety and maintenance by enabling more reliable and precise TPMS.

In recent years, the widespread adoption of image capture equipment and advances in image-based computer graphics have made the extraction and application of image structure information, such as gradient fields and lines, a key focus in the field of image processing. Nowadays, most mobile devices have relatively good shooting functions, so people’s demands for multimedia information processing are becoming stronger and stronger. In this study, an importance-driven picture composition method combining various rendering styles is proposed. The importance of each pixel in the picture is estimated based on image saliency or specified by user interaction. This method combines various rendering styles such as color desaturation, line rendering, edge preserving blur and contrast enhancement, and is suitable for image, video, 3D scene and even mixed reality scene. Highlighting the importance of image structure analysis, this study explores innovative methods to enhance the visual quality of augmented reality and stylization algorithms for images and videos. By delving into various rendering techniques, such as color desaturation and line rendering, it aims to significantly improve both the aesthetic appeal and functionality of digital media across diverse platforms, thereby meeting the growing demands for advanced multimedia processing.

With the increased operating time and usage frequency of hydroelectric equipment, monitoring and predicting its health state has become increasingly important. Traditional health assessment methods often rely on single measurement point data, which has problems such as insufficient model precision and poor real-time performance. These methods are difficult to reflect the overall health state of the equipment fully and cannot accurately capture the complex dynamic characteristics of the equipment, lacking real-time monitoring capabilities. This article proposed a comprehensive health prediction model for hydroelectric equipment based on the multimeasurement point output, which realized real-time monitoring and prediction of the health state of hydroelectric equipment, optimized maintenance strategies, and reduced maintenance difficulty. By collecting temperature signal data, vibration signal data, pressure signal data, and lubrication degree signal data from three systems including upper guide bearing, thrust guide bearing system, and water guide bearing in hydroelectric equipment, they were used as original sensor signal data. Data preprocessing is performed on original sensor signal data, including handling missing and outliers, stabilizing nonstationary time series data, and filtering temporal noise to address the impact of diverse types and large differences in data scales, achieving more accurate predictions. This article used sequential Bayesian method (SBM), hypersphere algorithm, and long short-term memory (LSTM) network to construct a prediction model. These algorithms had different advantages and applicable scenarios and could complement each other to improve the precision and robustness of prediction models. To further improve prediction precision, the model parameters were optimized through cross validation to avoid overfitting and improve model performance. By comparing and analyzing the predictive performance, error results, and real-time prediction performance before and after model optimization, it was concluded that the prediction model constructed by SBM, hypersphere algorithm, and LSTM network had an overall average improvement of 23.7% in the prediction precision of 12 parameters, including temperature, vibration frequency, pressure, and lubrication degree, for the optimized upper guide bearing, thrust guide bearing, and water guide bearing systems. The overall average error has decreased by 55.6%, and the real-time prediction performance has improved by 10.4%. The optimized model not only improves accuracy but also significantly enhances real-time performance, which can better adapt to real-time prediction needs in dynamic environments.

With the rapid development of modern society and economy, various types of science and technology have also achieved relatively rapid development, which has led to significant changes in all walks of life in modern society. This change not only makes modern society transition toward an information society but also makes people in modern society obtain more convenience. The construction industry has also achieved higher quality development with the development of science and technology, especially through the informatization reform of existing technical means in multiple processes such as building design and construction, which has saved a lot of manpower and material resources. The most important aspect in the field of architecture is the design work before construction, and the degree of refinement in this process also determines the merits of the building to a certain extent. Therefore, the field of architectural design has also received more attention from relevant researchers. At the same time, the further development of social economy in the new era also puts forward more requirements for architectural design in the construction industry, which urges researchers to conduct in-depth research on existing architectural design. At the same time, combining some emerging information technologies, a new architectural design mode with better structure and performance is proposed. Image processing technology mainly uses computer algorithms to collect images, thereby converting these images into digital signals that can be recognized by a computer, and then displaying them on a computer display. This image processing technology can also identify and extract information from images, thereby displaying the key information therein. This article mainly analyzes image processing techniques and some data analysis algorithms to obtain the feasibility of their application in visual information mining systems for architectural design. The contribution of this study is to propose a new architectural design visual information mining system based on image processing technology and data analysis algorithms, and conduct experimental studies to prove that it is superior to existing systems in terms of information extraction accuracy, processing speed, and user convenience. The experimental results show that compared with the existing architectural design data mining system, the new system has an average performance improvement of 25.9% in many aspects, which marks an important step forward in the field of architectural design informatization. In addition, this study also provides a new perspective for the construction industry, i.e., applying image processing and data analysis technology to the design process to improve design quality and efficiency.

In response to the deformation and displacement problems that exist during the construction and use of underground projects, this study develops an underground engineering robot measurement deformation monitoring system based on the Internet of Things (IoT) architecture, aiming to improve the accuracy and efficiency of underground engineering deformation monitoring. The system was designed with a multi-layer architecture, including a perception layer (real-time collection of deformation data) and a network layer (using a robot to carry a data acquisition terminal and transmit data to a cloud server through wireless communication technology), platform layer (using cloud computing and big data technology to store, process, and analyze collected data), and application layer (visualization platform, convenient operation, and analysis of data). Fifteen underground engineering projects were tested, with ten key monitoring points selected for each project. Multiple regression analysis was used to evaluate the monitoring accuracy and stability and to explore the relationship between monitoring data and geological conditions and environmental variables. The experimental results show that compared with traditional methods, this system has significant advantages in stress monitoring: the stress monitoring error of conventional methods is about 0.5 MPa, while the error of this system is only 0.01 MPa, a difference of 50 times. The system of the IoT significantly improved the detection efficiency compared with the traditional manual surveying and mapping method, and the underground engineering robot measurement deformation monitoring system based on the IoT architecture took only 4 h to monitor the small deformation. However, the traditional method took 15 h, i.e., the system based on the IoT was 11 h faster than the conventional method. The underground engineering robot measurement deformation monitoring system based on the IoT architecture has high reliability and operability, and provides strong support for the safety management of underground engineering.

Face attribute recognition is also widely used in human–computer interaction, train stations, and other fields because face attributes are rich in information. A convolutional neural network and distributed computing (DC)-based face recognition model is researched to enhance multitask face recognition accuracy and computational capacity. The model is also trained using a multitask learning method for face identity recognition, fatigue state recognition, age recognition, and gender recognition. The model is streamlined by DC to improve the computational efficiency of the model. To further raise the recognition accuracy of each task, the study combines the add feature fusion (FF) principle and concat FF principle to propose an improved face recognition model. The outcomes of the study revealed that the true acceptance rate value of the model reached 0.95, and the face identity recognition accuracy reached 100%. The fatigue state determination accuracy reached 99.12%. The average recognition time of a single photo was 354 ms. The model can quickly recognize face identity and face attribute information in a short time, with short time and high recognition accuracy. It is beneficial in several areas, including intelligent navigation and driving behavior analysis.

A fault localization and analysis method for Optical Power-Grade Ground Wire (OPGW) based on transformer and federated learning (FedL) in a cloud edge collaborative environment is proposed. First, based on the cloud edge collaboration architecture, a model framework for OPGW fault location is designed through the collaboration between the cloud center and edge computing. Then, by introducing FedL for model training at each OPGW edge sensor, only model parameters are exchanged without transmitting raw data, greatly reducing computational costs and network bandwidth requirements. Finally, the Transformer network was introduced into the model, which greatly improved the processing efficiency of fault data through parallel computing. The simulation experiment results show that the relative error, absolute error, and localization time of the proposed method for fault localization are the smallest on different datasets, with the lowest values being 0.78%, 0.0297 km, and 5.33 µs, respectively.

The goal of outlier detection is to pinpoint data points that exhibit notable deviations from the rest of the observed values. It has found successful application in numerous fields, including process inspection, anti-terrorist operations, and public security. However, the majority of existing outlier algorithms rely on methods involving filling or deleting missing data, with few directly addressing incomplete data. This article studies outlier detection for incomplete real-valued data based on class-consistent technology, rough set theory, and granular computing. First, a tolerance relation founded on class-consistent technology is presented to illustrate the similarity among information values within an incomplete real-valued information system (IRVIS). Then, the tolerance classes are established based on the tolerance relation and utilized for computing approximate accuracy and other metrics. Next, an outlier factor is defined, considering both the degree of outlierness and the weight function assigned to each object within an IRVIS, elucidating the uncertainty and degree of outlier. Finally, an outlier detection algorithm (ODIRG) for an IRVIS based on class-consistent technology, rough set theory, and granular computing is devised. Numerical experiments on seven UCI datasets are undertaken to evaluate the stability of the ODIRG algorithm. The proposed method is demonstrated to exhibit strong effectiveness and adaptability for categorical data when compared with five other algorithms. It is notable that for comprehensive comparison, precision, recall, F F 1-measure, and receiver operating characteristic curve are employed to delineate the benefits of the proposed approach.

Pulmonary diseases rank among the leading causes of mortality worldwide, underscoring the importance of early detection to enhance patient outcomes and reduce fatalities. Chest X-rays (CXRs) serve as a critical diagnostic tool for identifying lung diseases; however, many pulmonary conditions exhibit similar patterns on CXRs, making differentiation challenging even for experienced radiologists. Furthermore, a single CXR may reveal multiple coexisting diseases, such as pneumonia and pleural effusion. Moreover, most existing studies in this field have been constrained to a limited number of classes and datasets, leaving significant gaps in the classification of diverse pulmonary conditions, particularly in remote regions where access to skilled radiologists is scarce. Recently, vision transformers (ViTs) have become a key technique in deep learning, and they have been widely used to detect different diseases from CXRs. This study aimed to adapt and optimize the pre-trained ViT-B16 model to predict and distinguish between single and dual pulmonary diseases whose diagnoses relied only on CXRs. The goal was to discriminate eight classes of evidence, which consisted of individual diseases (COVID-19, pneumonia, pneumothorax, tuberculosis, and pleural effusion), dual diseases (pneumonia and pleural effusion as well as pneumothorax and pleural effusion), and normal status. The experimental results show that the proposed model distinguished more classes than previous methods, with an accuracy of 98.18%. To the best of our knowledge, this is the first study to attempt to discriminate eight classes of evidence concerning pulmonary diseases, including both individual and dual diseases.

Problem In continuous deep learning scenarios, where the model is required to learn new tasks without losing knowledge from prior tasks, catastrophic forgetting is a significant limitation of continuous deep learning models. This problem has the highest significance in the field of medical image analysis, as it is imperative to ensure the reliability of diagnostics and prediction through consistent and accurate classification. This study solves the challenge of catastrophic amnesia in deep learning by utilizing the conjunction of Convolutional Neural Networks (CNNs) with the Elastic Weight Consolidation (EWC) called CNN-EWC model. Aims This research seeks to develop a continuous learning pipeline for classifying histopathology images of lung cancer into lung benign tissue ( n ), lung adenocarcinoma (ACA), and lung squamous cell carcinoma (SCC). A dataset of 15,000 images equally across these three classes is used to comprehensively evaluate the proposed method in a realistic medical diagnostic scenario. Methods The suggested CNN-EWC method mitigates catastrophic forgetting in continuous deep learning by safeguarding critical parameters across tasks. The model is trained successively on three tasks, utilizing EWC to inhibit substantial alterations to essential parameters. This ensures a preservation of knowledge from previous tasks while accommodating new tasks. The methodology preserves elevated classification precision in medical diagnostics by balancing stability and adaptability. Results The CNN-EWC model experimental results achieved testing accuracy from 98 to 99.6%. Notwithstanding a diversity in dataset size (3,000, 4,500, and 6,000 images), the model performed excellently for lung n, lung ACA, and SCC, its good statistics hold true for all three diseases tested on the testing dataset. External tests were conducted on (3–1,500) images to verify the accuracy of the model’s classification and prediction, and the accuracy reached 100%, which indicates that the problem of catastrophic forgetting that occurs in continuing education has been addressed. Conclusions The CNN-EWC model has proven its effectiveness in continuous training in terms of accuracy, speed, saving time, and correct classification of images as it solves catastrophic forgetting but also maintains very high classification performance. In medical imaging, the simple switching logic effectively solves several simulations of catastrophic forgetting by avoiding back propagation for large numbers of images or by direct calculation.

With the expansion of the Internet, network functions and requirements are becoming more varied. To solve this problem, cloud computing adopts virtualization technology and refers to virtual network function migration to improve network adaptability. However, issues such as resource overload and inefficient migration still exist. Given this, the article proposes a bandwidth-aware virtual network migration algorithm. It combines the risk avoidance simulation learning algorithm with the good and bad distance solution method and applies it to the cloud computing virtual environment. The algorithm monitors bandwidth usage in real-time, identifying potential bottlenecks and considering both resource utilization and bandwidth availability when determining whether to migrate virtual machines. This approach avoids performance degradation due to bandwidth limits. The results showed that the algorithm outperformed others, achieving optimal performance in just 102 iterations and excelling in resource occupancy, utilization, load balancing, migration costs, average migration quantity, latency, and task completion time. This has enabled cloud providers to manage data center network resources in a more efficient manner, thereby enhancing operational efficiency, reducing costs, and delivering more stable and reliable cloud services, which hold significant practical and market value.

Nowadays, anomaly detection in streaming data has gained considerable attention due to the exponential growth in the data gathered by Internet of Things applications. Analyzing and processing vast data volumes requires a system capable of working in real-time. Moreover, obtaining labeled data for supervised learning is challenging, as it requires human involvement, is time-consuming, and costly. A promising direction is implementing evolving spiking neural networks (eSNN), which can be updated whenever new data becomes available without re-training previous samples. However, eSNN encounters significant challenges when it comes to manually tuning its hyperparameter values. As such, this work covers the current research gap by suggesting a novel method to optimize the hyperparameters of eSNN called online evolving spiking neural networks-artificial bee colony (OeSNN-ABC). Multiple scenarios have been utilized to evaluate the proposed method using two benchmark datasets: the Numenta anomaly benchmark (NAB) and the Yahoo Webscope using different criteria. Further validation was provided by comparing the proposed OeSNN-ABC against five well-known optimization algorithms: particle swarm optimization, grey wolf optimization, flower pollination algorithm, whale optimization algorithm, and grid search, alongside other classifiers such as random forest, support vector machine, and k-nearest neighbor. The findings revealed that OeSNN-ABC had the best performance among all compared optimization algorithms and classifiers, outperformed prior anomaly detection techniques for the NAB dataset, and achieved competitive results for the Yahoo Webscope dataset.

Problem – Diagnosing early-stage oral squamous cell carcinoma (OSCC) is necessary for patient treatment and survival. The staining normalization and augmentation of histopathological images can play an important role in improving the accuracy of early oral cancer (OC) detection, especially with imbalanced and limited datasets, which is a difficult manual task because it is time-consuming, requires effort, and is subject to variation among different pathologists. Applying these methods can enhance the performance of deep learning models used for OSCC detection. Aim – This study addresses the aforementioned challenges by showing how the random stain normalization and augmentation (RandStainNA) technique can be applied to datasets as a preprocessing step along with transfer learning models to categorize histopathological images of OC into three classes. Methods – The performance of three models – ResNet-50, VGG-19, and an ensemble model – in image classification is compared. Using fivefold cross-validation for training, our framework compares two main tasks. The best results were achieved on the task in which the models were pretrained using patches, and the RandStainNA technique was applied to the images as a preprocessing step. We trained the models to extract features from the relevant images and then determine the final classifications. Results – Our proposed framework achieved the best results in the early diagnosis of OC using the NDB-UFES dataset and the ResNet-50 model, demonstrating 73.33% balanced accuracy, 74.68% precision, and 74.35% recall, as well as a 92.21% area under the curve. Conclusion – Our proposed framework was effective in improving the accuracy of early OC detection, especially when working with small and imbalanced datasets. In addition, this method may contribute to enhancing the generalizability of models and using them in diverse laboratories.

One critical research gap that this study fills is artificial intelligence (AI) and machine learning applications that predict equity market index total returns using long-term prediction horizons, and by experimenting with Extreme Gradient Boosting (XGBoost). The presented models achieved significantly higher accuracy rates than the majority class rate, and they obtained better predictive scores in all metrics than the logistic regression model. The best-performing model had a 100% accuracy rate when negative returns were predicted with a p -value of 0.05121. The evidence from this study suggests that XGBoost, a neglected algorithm in the literature, can enable more empirically informed long-term portfolio management decisions regarding overall equity exposure. Moreover, a literature contribution of this study is a refined methodological approach for prospective studies when implementing binary classifiers of prospective stock market returns for enhanced real-life economic utility for investors. The constructed models generated probabilities of whether the S&P 500 will have positive or negative total returns, including dividend payouts, in the subsequent 12 months. The predictive metrics of these models were evaluated against traditional logit models and whether the accuracy rates statistically significantly exceeded the majority class rate.

In South Africa, it is a common practice for people to leave their vehicles beside the road when traveling long distances for a short comfort break. This practice might increase human encounters with wildlife, threatening their security and safety. Here, we intend to create awareness about wildlife fencing, using drone technology and computer vision algorithms to recognize and detect wildlife fences and associated features. We collected data at Amakhala and Lalibela private game reserves located in the Eastern Cape province of South Africa. We used wildlife electric fence data containing single and double fences for the classification task. Additionally, we used aerial and still annotated images extracted from the drone and still cameras for the segmentation and detection tasks. The model training results from the drone camera outperformed those from the still camera. Generally, poor model performance is attributed to (1) over-decompression of images and (2) the ability of drone cameras to capture more details on images for the machine learning model to learn as compared to still cameras that capture only the front view of the wildlife fence. We argue that our model can be deployed on client-edge devices to inform people about the presence and significance of wildlife fencing, which minimizes human encounters with wildlife, thereby mitigating wildlife–vehicle collisions.

Traveling salesman problem (TSP) is a typical combinatorial optimization problem which is regarded as an NP (nondeterministic polynomial)-hard problem. When the TSP in a large scale are solved by traditional optimization methods, the computation is large, the iteration time is long, and the result may not be optimal. The genetic algorithm is a highly favored optimization algorithm, but due to the lack of diversity in population and genetic individuals, it also has the same limitations when solving large-scale TSPs. In order to improve the solving speed and solution accuracy of TSPs, an adaptive genetic algorithm with double populations (DPAGA) is proposed. A dominant population and an auxiliary population are introduced in the algorithm. The algorithm distinguishes genetic individuals into male and female individuals, and only heterosexual individuals can perform crossover operations. The dominant population and auxiliary population execute different adaptive crossover and mutation strategies during the evolution process. The DPAGA is used to solve TSPs; the experimental results prove that it is completely feasible, and its convergence speed and solution quality have been improved. When solving the TSPs with scales of 20, 30 and 50 cities, the optimal path lengths of DPAGA were reduced by the maximum of 0.2491, 9.1071 and 15.7086, respectively, compared with other optimization algorithms.

Stroke ranks as the second leading cause of death worldwide and is a major contributor to disability. Researchers have proposed various applications to assist in the rehabilitation of stroke patients, with brain-computer interfaces (BCIs) utilizing electroencephalograms (EEGs) showing particularly promising outcomes. However, the most challenging aspect of using BCI methods is effectively extracting and selecting the most significant features from the vast amount of EEG data available. This article addresses this problem by presenting an innovative optimization-based approach to the channel selection problem, employing a novel binary equilibrium optimizer (EO) as an optimization technique to identify the most relevant EEG channels. This method significantly enhances the accuracy of stroke patient rehabilitation outcomes while reducing computational complexity, avoiding overfitting, and minimizing user discomfort during clinical use. During the preprocessing step, conventional filters and the AICA-WT (automatic independent component analysis with wavelet transform) denoising technique are employed. Attributes are then computed for the time, entropy, and frequency domains. The EO algorithm is utilized to represent the EEG channel selection problem, transforming the features of each individual EEG into binary values and subsequently applying a k-nearest neighbor classifier technique to determine the accuracy rate. The proposed method demonstrates superior performance, achieving the highest accuracy rate of 99% with the HFD features, compared to the existing methods. In general, the proposed study provides a more reliable strategy by identifying a subject-specific reduced set of relevant electrodes and establishes a new benchmark in the field of stroke rehabilitation, highlighting the quality and potential impact of this work.

Intelligent manufacturing (IM) is an advanced model that integrates traditional manufacturing with artificial intelligence to achieve automation, intelligence, and flexibility throughout the production process. Chinese enterprises exhibit a relatively low overall innovation efficiency, with a central focus on IM as a powerhouse strategy. The developmental status of IM is pivotal, as it directly affects the quality standards of China’s manufacturing industry. This study employed machine learning techniques to examine whether IM can enhance the innovation efficiency of Chinese enterprises while also exploring the roles of government subsidies and intellectual capital in this process. The empirical findings are as follows. (1) Among Chinese listed companies, a higher level of IM significantly improved innovation efficiency, a result confirmed by robustness tests. (2) Mechanism analysis revealed that IM enhanced innovation efficiency primarily by securing government resource support and fostering intellectual capital. (3) Moderation analysis indicated that the positive correlation between IM and innovation efficiency weakened as the level of financial development increased, whereas intensified market competition strengthened this correlation. This study could provide valuable insights for Chinese manufacturing enterprises aiming to enhance their intelligence levels and promote innovation transformation and upgrading.

The increasing demand for operational efficiency and data integrity has led enterprises to prioritize the digital transformation of internal workflows. This is done through automation of document-related business processes. This study proposes an intelligent enterprise system that integrates artificial intelligence and big data technologies for the automated generation, validation, and approval of business documents. The motivation behind this work derives from the need to reduce human error, enhance accuracy, and accelerate document turnaround times in enterprise environments. The authors employ large language models to automatically generate document templates and a fine-tuned bidirectional encoder representations from transformer-based classifier for validating document content. Big data tools such as Apache Spark are used for processing and cleaning large volumes of enterprise documents. Additionally, low-confidence predictions are handled through a human-in-the-loop mechanism to ensure high reliability. The research process involves system design, data collection from over 9,000 real enterprise documents, model training, and integration into a business process management system. Experimental results show that the proposed approach improves document processing efficiency while maintaining data quality. This article presents a unified framework and implementation methodology that can be adapted for broader enterprise automation needs. The authors use data from thermal grid enterprises as an example for testing the developed intelligent models.

Accurate bankruptcy prediction is essential for financial stability, risk management, and decision making. Traditional statistical models cannot capture complex nonlinear relationships in financial data; hence, it is essential to use advanced deep learning techniques. This study proposes a hybrid long short-term memory (LSTM) model, optimized using the harmony search algorithm (HSA) for enhanced predictability. A financial dataset of Polish companies from the emerging markets information service was utilized, which consisted of financial indicators spanning several years. For feature selection, principal component analysis was employed to achieve dimensionality reduction without loss of pertinent financial data. The new HSA-LSTM model was compared with benchmark classifiers such as fuzzy logic neural network, kernel extreme learning machine, extreme learning machine, and support vector machine in terms of key performance measures. The HSA-LSTM model outperformed all of the baselines with 90.80% accuracy, 89.46% precision, 90.22% recall, and an F -score of 90.54%. Statistical verification using ANOVA and Tukey’s HSD test confirmed that the improvement was significant ( p < 0.05). These findings highlight deep learning with hyperparameter optimization for financial distress prediction. This study contributes to early bankruptcy prediction and financial risk modeling and offers a high-accuracy, scalable prediction model. Future research could explore additional metaheuristic optimization techniques, explainable AI (XAI), and macroeconomic indicators to further enhance predictive accuracy.

This study proposes a machine learning (ML) framework to overcome emotion recognition challenges in classroom environments, where high facial expression similarity and complex postural backgrounds hinder the accurate analysis. First, the histogram of oriented gradients (HOGs) was used to extract seven facial expression features. Second, comparative analysis of six ML algorithms identified support vector machine (SVM) as the optimal classifier. Third, grid search with cross-validation enhanced SVM’s recognition performance by 13.9% accuracy, 11.3% precision, and 13.8% recall improvement. Fourth, students’ facial expressions were recognized using HOG and the optimized SVM during eight course tasks. A classroom teaching effect evaluation model was constructed to predict students’ learning concentration scores according to the positive degree of different facial expressions. Absolute error, relative error, scatter, and violin plots all demonstrate that the predicted concentration score is strongly linearly correlated with actual mission score and final grade; mean absolute errors were 1.95 and 3.3, while mean relative errors were 2.57 and 4.42%, respectively. This study provides a reliable new method for intervening in students’ learning concentration in advance and fostering the quality of classroom teaching.

In this study, an artificial intelligence (AI) assistant is developed specifically for quality assurance (QA) tasks in response to the software industry’s growing demand for better QA solutions. Traditional QA methods are labor-intensive and prone to human error, even when they function properly. Utilizing the most recent advancements in natural language processing and AI, this AI assistant maximizes output and reliability by automating and optimizing QA processes. Using Rasa technology, the AI assistant aims to revolutionize QA by offering testers comprehensive support, including question-answering, contextual recommendations, and expanding users’ knowledge bases. It can reply to queries using both words and images. They are minimizing the amount of time spent searching for answers and raising user engagement. Notwithstanding AI’s potential to improve quality control, challenges remain, particularly in the area of human-like language production and comprehension. In today’s industry, professional QA tester training is a big worry. Due to the conventional training method’s substantial dependence on senior testers to educate and assist trainees, only two or three trainees may typically obtain appropriate instruction at a time from a single senior QA tester. This limitation affects the scalability of QA training programmers. With the new approach, QA training can be made much more productive and scalable since a single senior QA tester can now educate over 20 new hires at once.

A measurement data anomaly analysis method based on similarity metric and improved deep reinforcement learning under the power Internet of Things (PIoT) architecture is proposed. First, a data anomaly analysis architecture based on the PIoT is built, and high-quality data processing is achieved through cloud-edge collaboration. Then, on the edge side, a kernel function is used to optimize the fuzzy C clustering algorithm to obtain data feature curves and possible abnormal samples are screened out through similarity metric methods for further analysis. Finally, a deep Q network (DQN) combined with a long-short term memory network (LSTM) is deployed in the cloud center, and data anomaly analysis results are obtained through training and analysis of the LSTM-DQN network. Based on the selected data samples, experimental analysis is conducted on the proposed method, and the results show that its analysis accuracy, response time, and stability are 96.07%, 220 ms, and 0.02, respectively, which can quickly and accurately analyze abnormal measurement data.

Skin cancer especially when detected early can be easily treated, but its diagnosis is complicated by the minimal difference in the appearance of early lesions and the requirement of a precise diagnostic technique. The goal of this intensive literature review is to evaluate the progressive enhancements of deep learning (DL) and machine learning (ML) methods for transferring early-stage skin cancer identification in terms of accuracy and in terms of usability for real-world clinical applications. By using support vector machines, convolutional neural networks, and ensemble methods, we assess the performance of such algorithms in the classification and segmentation of skin lesions within various datasets. The challenges outlined in the review include the following: first, sparsity of data, second, variation in the looks in lesions, and third, imbalance of data within classes. Furthermore, issues that are still open to investigation are also presented, including the restricted number of algorithms for which the developed DL/ML models can be interpretable and the variability of the results assessment criteria used in different investigations. We then propose possible approaches to these issues such as data augments, multimodal learning, and the inclusion of explainable artificial intelligence approaches. The strengths of the present study consist of a comprehensive review of the limitations of contemporary methodologies and recommendations for future research on DL/ML-based systems for the early diagnosis of skin cancer. This research aims to highlight the best techniques and identify areas for future improvement. The study highlighted the key challenges of evaluating skin lesion segmentation and classification techniques, for instance, small sample size dataset, or selective and random image acquisition or even racial prejudice.

The liver, a complex and important organ in the human body, is crucial to many physiological processes. For the diagnosis and ongoing monitoring of a wide spectrum of liver diseases, an accurate segmentation of the liver from medical imaging is essential. The importance of liver segmentation in clinical practice is examined in this research, along with the difficulties in attaining accurate segmentation masks, particularly when working with small structures and precise details. This study investigates the performance of ten well-known U-Net models, including Vanilla U-Net, Attention U-Net, V-Net, U-Net 3+, R2U-Net, U 2 {{\rm{U}}}^{2} -Net, U-Net++, Res U-Net, Swin-U-Net, and Trans-U-Net. These variations have become optimal approaches to liver segmentation, each providing certain benefits and addressing particular difficulties. We have conducted this research on computed tomography scan images from three standard datasets, namely, 3DIRCADb, CHAOS, and LiTS datasets. The U-Net architecture has become a mainstay in contemporary research on medical picture segmentation due to its success in preserving contextual information and capturing fine features. The structural and functional characteristics that help it perform well on liver segmentation tasks even with scant annotated data are well highlighted in this study. The code and additional results can be found in the Github https://github.com/akalder/ComparativeStudyLiverSegmentation.

The enhanced use of artificial intelligence (AI) in organizations has changed and revolutionized the approaches to solving problems, processing information, and decision making. While the algorithms turned out to be highly effective, AI systems faced adversarial attacks, which can be described as slight alterations of inputs that would fool an AI algorithm. These attacks remain major challenges to the dependability and protection of AI systems and thus the need to develop stable and flexible protection strategies and procedures. The aim of this article is to discuss the existing trends in adversarial attack techniques and protection mechanisms. To this end, papers, exact match, and systematically applied operative inclusion/exclusion criteria pertinent to protection strategies against adversarial attacks in multiple databases were incorporated and used. Specifically, 1988 papers were retrieved from Web of Science, IEEE Explore, and Science Direct, which were published between January 1, 2021, and July 1, 2024, where we used 51 of the identified journal articles for the quantitative synthesis in the final stage. Thus, the protection taxonomy, which resulted from our analysis, discusses the motivation, and best practices in relation to the threats in question. The taxonomy also describes challenges and suggests other ideas on how to improve the robustness of adversarial attack systems. Not only this study is a response to gaps in the literature but it also presents the reader with a map for further studies. It is necessary to draw attention to the fact that an objective criterion must be introduced to measure the degree of defense, collaboration with researchers of other fields, and the necessity to consider the ethical implications of the created defense mechanisms. Our results shall assist industry practitioners, researchers, and policymakers in designing an optimal AI security that can protect AI systems against dynamic adversary strategies. This review provides a reference of entry to the topic of AI security and the challenges that may be encountered together with the measures that can be taken to forward the studies.

The concept of cloud computing has completely changed how computational resources are delivered and used. By enabling on-demand access to collective computing resources through the internet. While this technological shift offers unparalleled flexibility, it also brings considerable challenges, especially in scheduling and resource allocation, particularly when optimizing multiple objectives in a dynamic environment. Efficient allocation and scheduling of resources are critical in cloud computing, as they directly impact system performance, resource utilization, and cost efficiency in dynamic and heterogeneous conditions. Existing approaches often face difficulties in balancing conflicting objectives, such as reducing task completion time while staying within budget constraints or minimizing energy consumption while maximizing resource utilization. As a result, many solutions fall short of optimal performance, leading to increased costs and degraded performance. This systematic literature review (SLR) focuses on research conducted between 2019 and 2023 on scheduling and resource allocation in cloud environment. Following preferred reporting items for systematic reviews and meta-analyses guidelines, the review ensures a transparent and replicable process by employing systematic inclusion criteria and minimizing bias. The review explores key concepts in resource management and classifies existing strategies into mathematical, heuristic, and hyper-heuristic approaches. It evaluates popular algorithms designed to optimize key metrics such as energy consumption, resource utilization, cost reduction, makespan minimization, and performance satisfaction. Through a comparative analysis, the SLR discusses the strengths and limitations of various resource management schemes and identifies emerging trends. It underscores a steady growth in research within this field, emphasizing the importance of developing efficient allocation strategies to address the complexities of modern cloud systems. The findings provide a comprehensive overview of current methodologies and pave the way for future research aimed at tackling unresolved challenges in cloud computing resource management. This work serves as a valuable resource for practitioners and academics seeking to optimize scheduling and allocation in dynamic cloud environments, contributing to advancements in resource management strategies of cloud computing.

The Healthcare Internet of Things (H-IoT) has made significant strides in transforming remote diagnostics, real-time monitoring, and data collection. However, achieving low latency remains a challenge, even for systems that critically depend on it, such as emergency response systems and remote surgeries. Fog computing (FC) emerges as a promising solution to mitigate latency issues by bringing computational resources closer to edge devices. This review provides a comprehensive analysis of latency optimization approaches within the H-IoT, framed within the context of FC. The study focuses on architectural enhancements, data communication methods, resource management strategies, and load distribution techniques aimed at reducing latency. Additionally, it evaluates existing challenges related to energy efficiency and trade-offs, security considerations, and scalability potential. Based on an analysis of recent advancements in the field, this study outlines potential development trends and future research directions for building adaptive, fast, and secure H-IoT systems. The study reaffirms the pivotal role of FC in shaping the future of healthcare services, particularly those with stringent low-latency requirements.