Faculté des mathématiques et de l'informatique
Permanent URI for this communityhdl:123456789/17
Browse
5 results
Search Results
Item Intelligent algorithms for feature selection in supervised and unsupervised classification(University of Mohamed El Bachir El Ibrahimi - Bordj Bou Arréridj, 2026) Khaoula Zineb LegouiThe growing availability of high-dimensional datasets across various domains has made feature selection a critical step in machine learning pipelines, as reducing irrelevant or redundant features enhances model interpretability and generalization. Due to the combinatorial nature of feature selection, traditional meth ods often lack the scalability and adaptability required for real-world problems. In response, this thesis investigates the application of intelligent metaheuristic algorithms to feature selection in both supervised and unsupervised learning settings. First, a comparative analysis of the Equilibrium Optimizer (EO) and Henry Gas Solubility Optimization (HGSO) algorithms is conducted for supervised classification tasks. Both algorithms are adapted to a binary feature space and evaluated on benchmark datasets using classification accuracy and feature reduction as performance criteria, highlighting their respective strengths and motivat ing a hybrid approach. Consequently, this thesis proposes HGSOEO, a hybrid algorithm that integrates the complementary exploration and exploitation capabilities of HGSO and EO. The proposed HGSOEO algo rithm is evaluated on the Twitter Spam Detection dataset and demonstrates superior performance in terms of classification accuracy and the number of selected features when compared to conventional metaheuristic and classical feature selection methods. Furthermore, the application of EO is extended to feature selection for clustering tasks, where labeled data are unavailable, by employing clustering validity criteria such as the Adjusted Rand Index (ARI) to guide the selection process. Experimental results across multiple datasets confirm the effectiveness and robustness of the proposed approaches. Overall, the findings of this thesis demonstrate that intelligent metaheuristic algorithms provide efficient and scalable solutions to the feature selection problem in both supervised and unsupervised learning contexts.Item Machine Learning for Misbehavior Detection in Next-Generation Vehicular Networks(university of bordj bou arreridj, 2025) MADI Ahmed Salah Eddine; MEKHFI BayaConnected vehicles have great potential to enhance road safety, reduce traffic congestion, and play a vital role in green engineering by reducing pollution and fuel consumption. By enabling more efficient traffic flow, eco-routing, and optimized driving behaviors, connected vehicles contribute to a cleaner environment. However, when a vehicle is compromised, it can pose a serious threat to the entire network due to the potential harm it can cause. One of the major challenges in vehicular networks is the detection of misbehaving vehicles, which should then be blacklisted or their certificates revoked. In this work, we propose a novel scheme that leverages machine learning to accurately detect and classify vehicle behavior, enabling effective identification and management of misbehaving vehicles. To assess the effectiveness of our approach, a comprehensive comparative analysis was performed. The results demonstrate that our model outperforms existing methods in accurately classifying vehicle behaviors, highlighting its potential for real-world deployment in securing vehicular networks.Item Classification des Signaux EEG basée sur l’Apprentissage automatique et l'apprentissage profond(university of bordj bou arreridj, 2025) Temhachet Rayane; Laichaoui YasminaApproximately 50 million people worldwide suffer from epilepsy, a chronic neurological disorder. The automatic detection of epileptic seizures from EEG (electroencephalogram) signals remains a major challenge for researchers. This study proposes a methodology for classifying EEG signals using machine learning techniques, including Support Vector Machines (SVM), k-Nearest Neighbors (KNN), the Naive Bayes (NB) algorithm, and a Convolutional Neural Network (CNN). To evaluate model performance, metrics such as Accuracy (AC), Sensitivity (SE), Specificity (SP), and Receiver Operating Characteristic (ROC) curves were used. The results demonstrate the effectiveness of the applied models, particularly the CNN, in the automated detection of epilepsy from EEG signals.Item Impact des techniques de prétraitement sur la performance des modèles de classification du diabète.(university of bordj bou arreridj, 2025) Dif marwa Mahmoud; Zineb GhezlaneDiabetes is a chronic disease for which early diagnosis is crucial to prevent serious com plications. In this work, we study the impact of preprocessing techniques on the performance of classification models applied to diabetes data. To this end, we use two medical datasets : the Pima Indians dataset and a local dataset from Iraq. We evaluate three classification algo rithms : logistic regression, support vector machines (SVM), and decision trees. We apply two normalization techniques (MinMaxScaler and StandardScaler) and three feature selection me thods (SelectKBest, GenericUnivariateSelect, SelectFromModel). The results, evaluated using cross-validation, show that a well-chosen preprocessing strategy significantly improves model accuracy, with varying performance depending on the nature of the data and the algorithm used.Item Fonctionnalités approfondies pour les systèmes de vérification Palmaire(university of bordj bou arreridj, 2024) - KACIMI lina; - TABET chaimaBiometrics is the automated identification of individuals based on their physical and behavioral characteristics. It helps provide certainty when interacting with familiar or unfamiliar people, authorizing the granting of specific rights or the denial of certain privileges. The underlying principle of biometrics is the assumption that each individual has unique physical and behavioral characteristics that distinguish them from others. Improving human identification techniques currently focuses on exploring new and emerging methods. This development is driven by growing security concerns and the emergence of tampering techniques. The goal is to leverage distinct parts of the human body that can be used for accurate identification, such as fingerprints, palm prints, iris and lips. However, many existing systems and methods suffer from slow processing or require expensive technical equipment. Palmprints have proven to be a promising biometric modality for personal identification due to their uniqueness and stability. This master's dissertation presents an in-depth study on the use of deep features for palm print identity verification systems. We have experimented with CNN models for pre-processing of TANTRIGGS, DOG methods and for feature extraction such as BSIF, GABOR. For classification, we used K-Nearest Neighbors (KNN), Support Vector Machines (SVM), ALMO.