Epilepsy is a prevalent neurological disorder affecting 50 million individuals worldwide and 1.2 million Americans. There exist millions of pediatric patients with intractable epilepsy, a condition in which seizures fail to come under control. The occurrence of seizures can result in physical injury, disorientation, unconsciousness, and additional symptoms that could impede children's ability to participate in everyday tasks. Predicting seizures can help parents and healthcare providers take precautions, prevent risky situations, and mentally prepare children to minimize anxiety and nervousness associated with the uncertainty of a seizure. This research proposes a novel and comprehensive framework to predict seizures in pediatric patients by evaluating machine learning algorithms on unimodal neuroimaging data consisting of electroencephalogram signals. The bandpass filtering and independent component analysis proved to be effective in reducing the noise and artifacts from the dataset. Various machine learning algorithms' performance is evaluated on important metrics such as accuracy, precision, specificity, sensitivity, F1 score and MCC. The results show that the deep learning algorithms are more successful in predicting seizures than logistic Regression, and k nearest neighbors. The recurrent neural network (RNN) gave the highest precision and F1 Score, long short-term memory (LSTM) outperformed RNN in accuracy and convolutional neural network (CNN) resulted in the highest Specificity. This research has significant implications for healthcare providers in proactively managing seizure occurrence in pediatric patients, potentially transforming clinical practices, and improving pediatric care.
Probabilistic self-organizing maps and Gaussian mixture models represent flexible and interpretable probabilistic models that address various machine learning needs, such as speech processing and compression. In this context, the probabilistic self-organization map (PRSOM), as an extension of the classical Kohonen self-organization map (SOM), estimates the density distributions of the data using a combination of normal distributions. However, the likelihood function in the normal mixture may exhibit unbounded features and spurious local maxima (degeneracy). To tackle this problem, we introduce specific constraints to the PRSOM model, based on Ingrassia’s approach. Finally, we provide an implementation of the proposed method and give a comparative evaluation of its performance through numerical experiments.
Artificial intelligence (AI) has the potential to have an impact on every aspect of a business, including the purchasing division. This article analyzes how AI is used in the fulfillment of the purchasing function in enterprise resource planning (ERP) systems, especially SAP editor (Systems, Applications, and Products in Data Processing), and its impact on the system's performance. First, a review is undertaken of how AI is used in different fields in companies. Thereafter, the paper follows a deep research to collect the information on SAP editor from any sources such as industry reports, article ad information systems site webs. The results indicate that the application of AI can be integrated with the purchasing function of ERP systems almost throughout their entire lifecycle, helping overcome many problems that ERP systems are not able to automate and streamline.
As online learning evolves rapidly and discussion forums become increasingly important in educational environments, this study analyzes the impact of participation in discussion forums on the learning performance of future computer science teachers. Data were collected to assess their level of engagement and academic outcomes. The methodology included the use of natural language processing techniques, such as the Bidirectional Encoder Representations from Transformers (BERT) model, and Latent Semantic Analysis (LSA) to analyze the emotional state and relevance of their contributions. Preliminary results indicate a positive correlation between active participation and their learning performance. Additionally, this study makes a significant contribution by analyzing the correlation between various participation factors and learner performance. These findings highlight the importance of discussion forums as interactive educational tools to enhance learning and promote learner engagement. Important implications are also presented in relation to the obtained results.
In this paper, we will focus on how highly variable characteristics impact assembly line balance. Mass customization makes balancing the production line more difficult. Typical assembly lines are designed to produce a single product or a narrow range of comparable items in high quantities, the goal of this research is to determine the essential components that have a substantial influence on the balance of the assembly line. To do this, the research proposes a 3D matrix interpretation and statistical methods using the correlation study with the MATLAB tool. The objective is to discover the highly changeable parameters and the interaction between each parameter that impact the line's balance. The results of this approach show that altering the piloting parameters of the assembly line has an influence on its balancing, and hence the line must be dynamically balanced. The proposed approach has been used in the automotive sector, notably in the wire harness manufacturing line.
As Morocco stands at the precipice of a transportation revolution driven by the surge of electric vehicles (EVs), the necessity for a modern, adaptive infrastructure becomes paramount. This North African nation looks to intertwine the burgeoning narrative of EVs with the momentum of digital transformation, aiming for an integrative approach to urban development and mobility.While cities across Morocco are eager to adapt, the country faces the dual challenge of optimizing vehicular flow for EVs and assuring prompt, efficient access to charging infrastructures. The national ambition, deeply rooted in Morocco’s progressive vision, is to transform its urban centers into paradigms of smart urbanism on the global stage.In our research, we employ the innovative capabilities of the Internet of Vehicles (IoV). A focal point of our study is the real-time simulation via SUMO (Simulation of Urban MObility), enriched by the data and control possibilities offered by the Traffic Control Interface (TraCI). Our proposed approach, grounded in an intelligent system, is designed to both refine road traffic management and enhance the connectivity between EVs and charging stations. Utilizing Agadir as our exemplar for the SUMO simulation, We emphasize the system’s commitment to seamless and effective interaction. This research not only underscores Morocco’s dedication to sustainable transportation but also positions it as an emerging frontrunner in smart city initiatives.
Connected environments collect data from millions of devices every second presenting valuable information to the community. However, data integrity faces critical research gaps especially when it comes to data obsolescence and its detection. Thinking of data obsolescence as an independent terminology, we define it as the state wherein data is no longer significant or effective with respect to the device, parent zone and/or the environment. Accordingly, this paper provides connected environments with a dimensional architecture that : (1) assess predefined quality metrics of each sensed data, (2) discovers interrelation between deployed devices using clustering techniques, and (3) identifies data obsolescence through three main layers; namely, the device, hosting zone, and/or environment level.
The diagnosis of retinal diseases using the vasculature of Fundus images has long been a focus of both ophthalmologists and medical research. Using computer-aided techniques to provide segmentation of blood vessels helps ophthalmologists in their clinical application routine. However, segmenting the vessels from retinal images is still a challenging task because of the variability in retinal vascular thickness, inhomogeneity of image intensity, and noise textures(both pathologies or bad pixels) that might hinder the quality of images. This work introduces a new technique of image processing based on Hessian filter to extract the blood vessel structure with a three-stage approach to filter Thicker, Thinner, and Neo Vessels. The proposed method is evaluated on three publicly available datasets namely Drive, Stare, and ChaseDB respectively achieving an average accuracy of .93%, .94%, .95 % and then applied on our proper dataset of 350 images classified into 7 classes according to specific challenges. Manual validation by a team of four ophthalmologists provides useful insights into each class of images. Finally, we aim at providing an annotated database of fundus images which will serve as a training dataset for Deep Learning techniques.
Despite opaque machine learning models outperform transparent models, users are still unable to comprehend and trust their outcomes.Therefore, Explainable Artificial Intelligence (XAI) becomes a prominent research area. The main objective of this discipline is providing new techniques, tools and models that explain how opaque machine learning models operate to give predictions or simply offer some information about system decisions.In this article, we present an experimental evaluation of explainable models used in XAI, namely: LIME, SHAP, ANCHOR and EBM. Those models are applied on the results of two opaque machine learning models: Random Forest and XGBoost. Our experimental evaluation covers various aspects for comparison including:•Examining the explainability coverage of methods.•Comparing explanations provided by each method.
Microgrid systems that use renewable energy sources are becoming more and more popular as strong backup power sources for manufacturing facilities, especially during outages when grid electricity is unavailable. Although its use as a backup energy system is acknowledged for improving resilience, its concurrent operation with the grid to allow non- emergency industrial processes is still unexplored. This study introduces a combined dynamic decision-making framework for optimal control of both the manufacturing and onsite power generation systems. We employ a Deep Reinforcement Learning (DRL) approach, specifically utilizing Deep Q-Learning and Double Deep Q-Learning models to estimate the value function of the decision-making model, originally formulated as a Markov Decision Process (MDP). Through a case study involving a manufacturing system with an onsite microgrid, we seek to validate our proposed model and solution strategy. This strategy is expected to reduce production costs overall while maintaining, if not improving, operational effectiveness even under non-emergency situations. Thus, our study makes important contributions to the continuing debates over energy conservation and sustainable manufacturing.
The distributed no-wait flow-shop manufacturing problem with sequence dependent setup time (DNWFSMP-SDST) is a challenging optimization problem that arises in various industries, including manufacturing and logistics. This problem involves scheduling a set of jobs on multiple machines in a distributed environment, where each job has a specific processing time and requires setup time between different machine transitions. The objective is to minimize the maximum tardiness of all jobs, which represents the maximum delay in completing any job beyond its desired due date. This study addresses the DNWFSMP-SDST optimization problem by employing two algorithms: Nawaz-Enscore-Ham (NEH) and Greedy Randomized Adaptive Search Procedure (GRASP). The NEH is a constructive heuristic that constructs an initial solution by iteratively inserting jobs into a partially ordered sequence. In contrast, GRASP combines a greedy randomized construction phase with a local search procedure to iteratively enhance the initial solution. The simulation study compares NEH and GRASP algorithms for solving DNWFSMP-SDST using bench-marking instance. The results show the pros and cons of each algorithm, helping practitioners make better decisions. Where the NEH algorithm outperform for GRASP technique solving the kind of problem. Overall, this research has enabled us to improve our methods for solving scheduling problems in distributed environments.
Explainable Artificial Intelligence (XAI) has emerged as an essential aspect of artificial intelligence (AI), aiming to impart transparency and interpretability to AI black-box models. With the recent rapid expansion of AI applications across diverse sectors, the need to explain and understand their outcomes becomes crucial, especially in critical domains. In this paper, we provide a comprehensive review of XAI techniques, emphasizing their methodologies, strengths, and potential limitations. Furthermore, we present a case study employing six model-agnostic XAI techniques, offering a comparative analysis of their effectiveness in explaining a black-box model related to a healthcare scenario. Our experiments not only showcase the applicability and distinctiveness of each technique but also provide insights to researchers and practitioners seeking sui table XAI methodologies for their projects. We conclude with a discussion on future research perspectives in the field of explainable AI.
The current e-health systems suffer from multiple difficulties due to their centralized system structure for managing Electronic Medical Records (EMR), this approach exposed security and privacy issues, leading to serious degradation of healthcare services and potentially compromising the confidentiality and integrity of private patient data. In this paper, we proposed a consistent architecture that adopts a decentralized design of an e-health monitoring system, incorporates Edge Computing, Internet of Medical Things, Hyperledger fabric network, and IPFS aims to guarantee a secure data exchange environment, while addressing the major existing e-health system's challenges in terms of performance and security aspects. Based on the existing centralized structure to manage medical data, the current structures reveal critical security vulnerabilities affecting personal data. By taking advantage of a distributed structure the proposed architecture provides a high-scale secure environment that facilitates exchanging medical data over a protected network, aligning with the STRIDE Threat model structure. In addition to embedded technologies including strong encryption algorithms and Role-based access control (RBAC) deployed to offer substantial security standards, and effectively mitigate various advanced cyber-attacks. This decentralized e-health design demonstrates an impactful improvement model evidenced by the experimental performance results and revealed a better security experience for sharing medical data among multiple stakeholders.
The requirements of optimizing path planning for the autonomous robot are in high demand in such industrial communities, especially in manufacturing and caring social support. The application of meta-heuristic methods in autonomous problems is considered because of their adaptivity and robustness. Ant Colony Optimization (ACO) is one of the researchers’ approaches because of its effectiveness in the ant community. However, some constraints are making this method less productive. Therefore, this paper is to generate a modified ACO combined with Fuzzy logic (ACOFL) to minimize its drawbacks and maximize the robustness of path planning. This study presents the role of ACO algorithms in Path Planning, especially in the complex aspect, through the theoretical background of ACO and the other combination with other algorithms. The improved mathematics model evaluation demonstrates the upgrade steps in the path-finding process. Simulation results shows that the modified ACO algorithm is effective for complex environment compare to standard ACO algorithm. Moreover, the comparison results are presented in this paper between the standard ACO and modified ACO algorithm.
Efficient Road sign recognition is key to improving road safety, navigation and to enhance driver assistance for an intelligent transportation system (ITS). However, achieving efficient road sign and traffic recognition is accompanied by many challenges, in our last few years, several deep learning algorithms have been deployed for responding to these limitations. With the exponential increase in the number of vehicles in Morocco, the need for accurate and real-time detection and classification of road signs has become essential. In this paper, we propose an advanced approach for recognizing Moroccan road signs utilizing the YOLOv8 (You Only Look Once) model that very known for its efficiency and effectiveness in object detection tasks. the model is trained and evaluated under a comprehensive dataset that comprising diverse Moroccan road signs. The test of YOLOv8 model give an encouraging result in different performance metrics such as accuracy, recall, precision and mAP which have respectively an average of 0.95, 0.94, 0.96 and 0.97.
Addressing the challenges of sustainable consumption in the energy sector, innovative approaches are required to motivate consumers behavior’s change. This article conducts a systematic literature review to study gamification applications in the energy sector to promote sustainable consumption. The objective is to evaluate gamified applications in various energy domains and identify the key gamification elements. Through this study, we observe an increase in the utilization of gamification in the energy sector over time, points, badges, feedback, and rewards are the widespread gamification elements, and those elements significantly change consumer behaviors, both in residential and workplace environments
In this paper, we present an innovative approach for the discovery of involutory maximum distance separable (MDS) matrices over finite fields F2q, derived from MDS self-dual codes, by employing a technique based on genetic algorithms. The significance of involutory MDS matrices lies in their unique properties, making them valuable in various applications, particularly in coding theory and cryptography. We propose a genetic algorithm-based method that efficiently searches for involutory MDS matrices, ensuring their self-duality and maximization of distances between code words. By leveraging the genetic algorithm’s ability to evolve solutions over generations, our approach automates the process of identifying optimal involutory MDS matrices. Through comprehensive experiments, we demonstrate the effectiveness of our method and also unveil essential insights into automorphism groups within MDS self-dual codes. These findings hold promise for practical applications and extend the horizons of knowledge in both coding theory and cryptographic systems.
Improving the heating, ventilation, and air conditioning (HVAC) system to minimize heating and ventilation consumption in large networks of controlled buildings has become a necessity nowadays. This attempt requires the development of a sophisticated machine learning (ML) model capable of properly predicting energy usage trends. This research focuses on the problem of predicting energy usage in a group of public buildings. Three unique models are used in the study: artificial neural networks, recurrent neural networks, and autoregressive neural networks. Given that the public sector consumes a significant portion of total consumption, accurately anticipating this usage becomes crucial in reaching energy conservation targets. We used data from an energy usage monitoring system placed in a business building in France for our experiment and we employ the aforementioned methodology, which depends on tree-based methodologies. We use measures such as RMSE, MSE, MAE, and R2 to analyze the performance of each technique. Our findings show promise in terms of accurately estimating energy use. In addition, we discovered the possibility of energy savings in one of the scenarios.
The precise prediction of petroleum product traffic holds vital significance for optimizing maritime operations and allocating resources efficiently within the industry. This research evaluates the performance of three distinct machine learning models, namely Extra Trees, K-Nearest Neighbors, and Decision Tree, and introduces an ensemble forecasting methodology to augment the accuracy of predicting traffic volume for petroleum products at an oil terminal. The investigation employs a two-year dataset derived from oil and gas tankers at Mohammedia port for analysis. Empirical findings reveal that the ensemble learner model surpasses the individual models in predicting traffic volume for petroleum products, demonstrating enhanced accuracy across both test and validation datasets.
Manufacturing flow-shop scheduling problem is a highly attractive optimization challenge in operations research. When there is limited space in a workshop or capacity queue of machine is null, a new problem called manufacturing blocking flow-shop scheduling problem (MBFSP) is investigated. This optimization problem is solved by considering that the job must be blocked during scheduling when the next machine is not available. The address this challenge effectively, we present two novel algorithms: the Iterative Local Search (ILS) with neighborhood solutions, and the Artificial Bee Colony (ABC) algorithm inspired by nature’s meta-heuristics. These cutting-edge approaches are designed to address the problem from distinct and innovative perspectives, with a primary focus on optimizing the overall quality of the solution. By embracing these advanced techniques, we can effectively expand the scope of potential outcomes, opening up new avenues for achieving superior results. The target is to decrease the total weighted flow time (TWFT) , which represents the duration of completing all tasks. We provide a mathematical statement of the MBFSP optimization case and propose two solution meta-heuristics. In order to comprehensively assess the effectiveness of the proposed algorithms, we conducted a rigorous comparative study. This study involved an extensive set of instances, carefully curated to encompass a wide range of scenarios with varying numbers of jobs and machines. By testing the algorithms across diverse instances, we aimed to obtain a holistic understanding of their performance. Simulation results show that the ABC algorithm overcame the ILS algorithm, with a hit ratio 88% over the set of test instances and a minimum mean relative deviation in percent of 0.0154%.
