http://univ-bejaia.dz/dspace/123456789/270 2026-04-08T21:05:58Z http://univ-bejaia.dz/dspace/123456789/25965 Title: Bio-inspired computation for bio-informatics problems. Authors: Dabba, Ali; Tari, Abdelkamel ; directeur de thèse Abstract: The field of bioinformatics opens up great opportunities to understand biological phenomena, which has attracted great interest from the scientific community in recent years. Consequently, there are many problems of bioinformatics, including multiple sequence alignment, protein structure prediction, construction of the phylogenetic tree and molecular docking, etc., which need the cooperation between biologists and computer scientists to be solved. This work addresses two problems: multiple sequence alignment and gene selection using bio-inspired algorithms. Firstly, we developed a method to solve the multiple sequence alignment problem, called a multi-objective artificial fish swarm algorithm (MOAFS), using the behaviors of artificial fish swarm algorithms, Pareto optimal set, and genetic operations. Secondly, we proposed an algorithm to solve the gene selection problem by using mutual information, moth flame optimization algorithm, and support vector machine with leave one out cross-validation (SVMLOOCV). It called the Mutual Information Maximization-modified Moth Flame Algorithm (MIM-mMFA) that consists of two simple phases. The thesis has processed a full test of the MOAFS on the BaliBASE 2.0 and BaliBASE 3.0 alignment benchmark datasets as well as the MIM-mMFA test on sixteen binary and multi-classes cancer gene expression datasets. Finally, we have given a deep insight into the performance of each algorithm. In addition, our proposed algorithms achieved competitive or better results than the wellestablished algorithms in the literature. Keywords: Bio-informatics; Bio-inspired Algorithms ; Multiple Sequence Alignment ; Artificial Fish Swarm Algorithm ; Gene Selection Genes Expression ; Microarray ; Cancer Classification ; Moth Flame Optimization Algorithm. Résume Le domaine de la bio-informatique offre de grandes possibilités de comprendre les phénomènes biologiques, ce qui a suscité un grand intérêt de la part de la communauté scientifique ces dernières années. Par conséquent, il existe de nombreux problèmes de bio-informatique, y compris l’alignement de séquences multiples, la prédiction de la structure des protéines, la construction de l’arbre phylogénétique et l’amarrage moléculaire, etc. qui nécessitent la coopération entre biologistes et informaticiens pour être résolus. Ce travail aborde deux problèmes: l’alignement de séquences multiples et la sélection de gènes à l’aide d’algorithmes bio-inspirés. Premièrement, nous avons développé une méthode pour résoudre le problème de l’alignement des séquences multiples, appelée algorithme d’essaim de poissons artificiels multi-objectifs (MOAFS), en utilisant les comportements des algorithmes d’essaim de poissons artificiels, l’ensemble Pareto-optimal, et les opérations génétiques. Deuxièmement, nous avons proposé un algorithme pour résoudre le problème de sélection de gènes en utilisant l’information mutuelle, l’algorithme d’optimisation de flamme de papillon de nuit, et le Machine à vecteurs de support avec leave-one-out cross-validation (SVM-LOOCV). Il a appelé Mutual Information Maximizationmodified Moth Flame Algorithm (MIM-mMFA) qui se compose en deux phases simples. La thèse a traité un test complet du MOAFS sur les ensembles de données de référence d’alignement BaliBASE 2.0 et BaliBASE 3.0 ainsi que le test MIM-mMFA sur seize ensembles de données du cancer binaires et multi-classes. Enfin, nous avons donné un aperçu approfondi des performances de chaque algorithme. De plus, nos algorithmes proposés ont obtenu des résultats compétitifs ou meilleurs que les algorithmes bien établis dans la littérature. Mots-clés: Bioinformatique ; Algorithmes bio-inspirés ; Alignement de séquences multiples ; Algorithme d’essaim de poissons artificiels ; Sélection des gènes ; Expression des gènes ; Puces à ADN ; Classification du cancer; Algorithme d’optimisation de la flamme papillon. Description: Option : Cloud Computing 2021-01-01T00:00:00Z http://univ-bejaia.dz/dspace/123456789/25964 Title: Mécanismes de formation de coalitions multi-agents avec externalités. Authors: Sklab, Youcef; Tari, Abdelkamel ; directeur de thèse Abstract: Formation de coalitions : Externalités dynamiques : Dépendances entre tâches* Description: Option : Cloud Computing 2021-01-01T00:00:00Z http://univ-bejaia.dz/dspace/123456789/25963 Title: Option : Réseaux et Systèmes distribués Authors: Djamila, Zamouche; Aissani, Soufiane ; directeur de thèse Abstract: Over the last decades, advancements in computing, electronics, and mechanical systems have been resulting in the development of transportation all over the world, which has been providing a lot of benefts for many aspects of human life. Intelligent Transportation Systems (ITSs) are advanced applications that aim to make the transportation infrastructures safer, more convenient, and smarter by using information that is shared among vehicles such as crash warning, sudden-brake warning, lane-change warning, and so on. Thus, such systems provide a wide variety of services including, but not limited to, trafc control, trafc management, passenger and road safety, and remote region connectivity. However, several challenges hampering the proper operation of these systems, such as extreme disturbances and they rely on several kinds of devices that can cause malfunctions. Moreover, vehicular communications are expected to be subject to severe breaches that a?ect the reliability of the exchanged information. Seeking to improve the safety and protect human life, in this thesis, we address these problems by providing improvements to the existing STIs. In particular, we have proposed an enhanced train-centric communication-based train control system for railway transportation that allows improving the quality and enhancing reliability of the train control. Moreover, we have proposed our second contribution that manifests itself in the proposition of a discordant safety messages detection strategy in connected vehicles environment that provides the vehicles with the ability to quickly and preemptively identify discordant messages and hence dealing against potential disturbances, while ensuring a trade-o? between the efciency and safety. The proposed mechanisms are evaluated through simulations in terms of important metrics. The obtained results highlight the promising performances of our proposals. Description: Option : Data Science 2022-01-01T00:00:00Z http://univ-bejaia.dz/dspace/123456789/25962 Title: Time synchronization in wireless senor and actuator networks. Authors: Boukhechem, Nadhir; Badache, Nadjib ; directeur de thèse Abstract: The purpose of time synchronization is to allow the different nodes' clocks in a network to get relatively close values at any moment. Currently, time synchronization is a fundamental problem in wireless sensor and actuator networks (WSANs). Indeed, many WSANs applications, including node localization, sleep schedule, and data aggregation, require accurate time synchronization to function properly. In this thesis, we propose two cluster-based time synchronization protocols for WSANs, namely Sensor and Actuator Networks Synchronization Protocol (SANSync), and Optimized Sensor and Actuator Networks Synchronization Protocol (OSANSync). These protocols, contrary to existing protocols, fully exploit the available resources of the actuators, particularly their large transmission range, to improve time synchronization accuracy. We also propose a heuristic-based method to select the ROOT node through which all the other nodes in the network are synchronized. The proposed method is fully distributed and can be easily integrated into time synchronization protocols to improve their performance. Description: Option : Réseaux et Systèmes distribués 2022-01-01T00:00:00Z