Browsing by Author "Touzout, Walid (Supervisor)"

Now showing 1 - 5 of 5

Design and implementation of an embedded AIOT-based home hospitalization system
(Université M’hamed Bougara de Boumerdes : Institut de Genie Electrique et Electronique, 2023) Rezrazi, Mohamed Seif Eddine; Djelti, Mohamed Amine; Touzout, Walid (Supervisor)
In recent years, the rise in chronic diseases among the elderly has emphasized the need for innovative healthcare solutions. The COVID-19 pandemic has further high- lighted the challenges faced by hospitals in accommodating a surge of patients while ensuring the safety of healthcare providers. To address these challenges, home hos- pitalization has emerged as a viable alternative. The Artificia lIntelligenc eo fThings (AIoT) based home hospitalization system integrates artificia lintelligence ,internet of things (IoT), sensors, and mobile/web applications to enable remote monitoring and management of patients’ health conditions. The system includes a hardware device with sensors for accurate data collection, a mobile application for patients to access health information and communicate with doctors, a web dashboard for doc- tors to manage patient data and provide personalized recommendations, and an AI model that analyze patient data and predict health conditions. The system employs a secure and reliable communication protocol for efficie ntda tatransmissio n.The primary objective of this system is to provide convenient and accessible healthcare services, particularly for the elderly and individuals with limited access to hospitals. By offerin gremot econsultation san dmonitoring ,th esyste mreduce sth enee dfor physical travel and ensures timely medical attention. The integration of AI and IoT technologies strengthens the system’s ability to support doctors in making informed decisions.
Edge-ML based nework intrusion detection system for IoT devices.
(Université M’hamed Bougara de Boumerdes : Institut de Genie Electrique et Electronique, 2024) Berkani, Lina; Khelifi, Cylia; Touzout, Walid (Supervisor)
In recent years, there has been a substantial proliferation in the use of the Internet of Things in a wide variety of domains, from providing new services and options in smart home applications to industrial IoT, automating healthcare, power grids and more. However, IoT networks are prone to security breaches due to the limited computational power and constrained resources of these devices, which cannot support traditional security mechanisms. This security concern is increasingly becoming a relevant research issue, for which a number of Network Intrusion Detection Systems (NIDSs) have been proposed. In this report, we develop and implement a practical machine learning based IoT network intrusion detection system that operates on low-end microcontrollers. The proposed system is deployed on edge which ensures a fast response to attacks targeting IoT devices, thanks to the decentralized data processing. Privacy of network users is also preserved as data is kept locally at the edge of the network. Two prototypes were proposed for the realization of this project, which are based on the Raspberry Pi and the ESP32 Microcontroller. The ESP32 based prototype is composed of three sub-systems, each utilizing an ESP32 MCU. Four distinct Machine Learning algorithms were explored to detect malicious from benign traffic, and recognize the type of attack, reaching up to an accuracy of 99.76% and an F1-score of 94.25% for tree-based models. A detailed evaluation and comparison of the models was conducted. The most accurate model was selected and then optimized to obtain a light weight and faster executable version, for an easier deployment on edge. The models were additionally tested on real-world data, by predicting the class label of previously unseen data from new Pcap files.
Image compression and decompression using deep convolutional autoencoder
(Université M’hamed Bougara de Boumerdes : Institut de Genie Electrique et Electronique, 2024) Zerabib, Abdellah; Hammi, Nadjemeddine; Touzout, Walid (Supervisor)
In the last decades, image compression has become a very important task since the Images contribute to a significan tamoun to finterne ttraff ic,demand ingefficient solutions to reduce storage and transmission costs while maintaining image quality. This thesis explores the use of Convolutional Autoencoders for image compression and decompression. Convolutional Autoencoders are employed to learn compact, meaningful representations of input images, which are then used to reconstruct the original images with minimal loss of quality. The model was tested using the Peak Signal-to-Noise Ratio (PSNR) and achieved good results, demonstrating the effectivenes so fth ecompression. Along with decompressing images, a network was trained to classify them bypassing the need for image reconstruction. The trained model compresses, classifies, and saves images efficiently, thus showing that compressed images canstill beaccurately categorized.
Optimized AI-based real-time state of charge (SOC) estimation of lithium-ion batteries.
(Université M’hamed Bougara de Boumerdes : Institut de Genie Electrique et Electronique, 2024) Bouchikh, Mohamed Amine; Tidjani, Mohamed Redha; Touzout, Walid (Supervisor)
Lithium-ion (Li-ion) batteries are highly valued for their ability to extend battery lifespan and enhance power energy density due to their chemical properties. The battery State of Charge (SOC) is a crucial parameter for monitoring battery health and estimating its lifespan, indicating how much longer the battery can be used and when it needs to be charged. Therefore, accurate SOC predictions are essential to prevent overcharging or over-discharging, and can be determined using conventional methods or data-driven approaches. In this report, we present a novel approach for SOC estimation of Li-ion batteries using optimized machine learning-based SOC estimation in both charging and discharging modes. The models are trained, tested, and optimized using a prognostic Li-ion battery dataset provided by the National Aeronautics and Space Administration (NASA) with five main inputs : load voltage, load current, measured voltage, measured current, and primarily battery temperature, the report is key for preferring data-driven approaches over conventional methods. Initially, nine different methods were evaluated : LASSO, K-Neighbor Regressor, CAT Boost Regressor, Extra Tree Regressor, Random Forest Regressor, XGB Regressor, Decision Tree Regressor, and Gradient Boosting Regressor, These methods were assessed in terms of their accuracy and the evaluation metrics R2, MAE, RMSE, with Four approaches showing promising results according to state-of-the-art applications namely : Extra Tree, XGB, DTR and Gradient Boot Regressors. Moreover, the novelty of this report involves hyperparameter tuning, including learning rate, maximum tree depth, number of trees and more, to find the optimal parameters for the three best models. Thus, GridSearchCV optimization method demonstrated significant improvement in terms of model evaluation metrics. Finally, the best approach (Extartree regressor) for charging and discharging was deployed onto an ESP32 microcontroller with an OLED interconnected with current, voltage, and temperature sensors for real-time battery SOC display and monitoring.
Service delivery management : multi-node path optimization
(Université M’hamed Bougara de Boumerdes : Institut de Genie Electrique et Electronique, 2023) El Hak Amer, Ouail Dhia; Sari, Mohamed El Amine; Touzout, Walid (Supervisor)
In an era marked by growing demands for efficie ntservi cedelive rya ndt heubiqui tyof advanced technologies, effectiv eservic emanagemen tha sbecom e apivota lconcer nfo rbusi- nesses across diverse sectors. The ability to optimize service routes not only drives customer satisfaction but also holds the promise of significan tcos tsavings .Thi sthesi sdelve sinto the heart of service delivery management, tackling the central challenge of Multi-node Path Optimization through the application and evaluation of two distinct yet complementary al- gorithms: the brute force algorithm and the heuristic algorithm. Rooted in the renowned Traveling Salesman Problem (TSP), this research endeavors to fin dth eshortes tpat hthat visits multiple destinations—a task of immense practical significance. The objectives of this study encompass a multifaceted exploration of these algorithms, en- compassing theoretical investigation, practical implementation, and comprehensive analysis. We aim to investigate the adaptability of both the brute force and heuristic algorithms to address Multi-node Path Optimization challenges, elucidate the computational intricacies associated with the TSP and its implications for service delivery management, craft a de- livery application using the versatile Flutter framework to provide a tangible manifestation of algorithmic efficienc y,a ndsubsequentl y,evalua tet hereal-wor ldperforman ce ofthe seal- gorithms in the realm of service delivery management. This research seeks to bridge the gap between theoretical optimization and practical application, offerin ginsight stha tcan empower businesses to make informed decisions in their pursuit of enhanced service delivery management. The finding so fthi sstud ypresen ta nopportunit yt orevolutioniz eservic edeliver yopera- tions across industries, offerin g apotentia lroa dma ptowar dincrease defficie ncy andcost- effectiveness .Throug h arigorou sexploratio no fMulti-nod ePat hOptimization ,thi sthesis contributes to the growing body of knowledge in the field ,offeri ngvaluab leperspectiv eson the intersection of theory and practice in service management.