Depuis 2004, les médias sociaux en ligne ont connu une croissance considérable. Ce développement rapide a eu des effets intéressants pour augmenter la connexionet l'échange d'informations entre les utilisateurs, mais certains effets négatifs sont également apparus, dont le nombre de faux comptes grandissant jour après jour.Les sockpuppets sont les multiples faux comptes créés par un même utilisateur. Ils sont à l'origine de plusieurs types de manipulations comme la création de faux comptes pour louer, défendre ou soutenir une personne ou une organisation, ou pour manipuler l'opinion publique. Dans cette thèse, nous présentons SocksCatch, un processus complet de détection et de groupage des sockpuppets composé de trois phases principales : la première phase a pour objectif la préparation du processus et le pré-traitement des données; la seconde phase a pour objectif la détection des comptes sockpuppets à l'aide d'algorithmes d'apprentissage automatique; la troisième phase a pour objectif le regroupement des comptes sockpuppets créés par un même utilisateur à l'aide d'algorithmes de détection de communautés. Ces phases sont déclinées en trois étapes : une étape "modèle" pour représenter les médias sociaux en ligne, où nous proposons un modèle général de médias sociaux dédié à la détection et au regroupement des sockpuppets ; une étape d'adaptation pour ajuster le processus à un média social spécifique, où nous instancions et évaluons le modèle SocksCatch sur un média social sélectionné ; et une étape en temps réel pour détecter et grouper les sockpuppets en ligne, où SocksCatch est déployé en ligne sur un média social sélectionné. Des expérimentations ont été réalisées sur l'étape d'adaptation en utilisant des données réelles extraites de Wikipédia anglais. Afin de trouver le meilleur algorithme d'apprentissage automatique pour la phase de détection de sockpuppet, les résultats de six algorithmes d'apprentissage automatique sont comparés. En outre, ils sont comparés à la littérature où les résultats de la comparaison montrent que notre proposition améliore la précision de la détection des sockpuppets. De plus, les résultats de cinq algorithmes de détection de communauté sont comparés pour la phase de regroupement de Sockpuppet, afin de trouver le meilleur algorithme de détection de communauté qui sera utilisé en temps réel. / Since 2004, online social medias have grown hugely. This fast development had interesting effects to increase the connection and information exchange between users, but some negative effects also appeared, including fake accounts number growing day after day. Sockpuppets are multiple fake accounts created by a same user. They are the source of several types of manipulation such as those created to praise, defend or support a person or an organization, or to manipulate public opinion. In this thesis, we present SocksCatch, a complete process to detect and group sockpuppets, which is composed of three main phases: the first phase objective is the process preparation and data pre-processing; the second phase objective is the detection of the sockpuppet accounts using machine learning algorithms; the third phase objective is the grouping of sockpuppet accounts created by a same user using community detection algorithms. These phases are declined in three stages: a model stage to represent online social medias, where we propose a general model of social media dedicated to the detection and grouping of sockpuppets; an adaptation stage to adjust the process to a particular social media, where we instantiate and evaluate the SocksCatch model on a selected social media; and a real-time stage to detect and group the sockpuppets online, where SocksCatch is deployed online on a selected social media. Experiments have been performed on the adaptation stage using real data crawled from English Wikipedia. In order to find the best machine learning algorithm for sockpuppet's detection phase, the results of six machine learning algorithms are compared. In addition, they are compared with the literature, and the results show that our proposition improves the accuracy of the detection of sockpuppets. Furthermore, the results of five community detection algorithms are compared for sockpuppet's grouping phase, in order to find the best community detecton algorithm that will be used in real-time stage.

<p dir="ltr">Pavement friction is fundamental to the safety of road networks. A precise assessment of friction levels is essential for the strategic development of maintenance practices and policies by state highway agencies. Typically, assessments of pavement friction have been conducted individually, focusing on particular segments of roadways. Nevertheless, this approach does not offer a thorough evaluation of roadway friction conditions at the network level. This study combines the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and the Gaussian Mixture Model (GMM) to evaluate the ratings of pavement friction throughout the entire state’s road system. A dataset oriented towards safety, serving as input for clustering models across various data dimensions, has been established. Through comparative and statistical analyses, six friction performance ratings have been identified and subsequently validated. The findings not only facilitate a deeper comprehension of the interrelations among friction levels, crash impact, and additional factors impacting safety, but also provide substantial insights for the advancement of road safety, management, and development.</p><p dir="ltr">Pavement markings play an essential role in regulating traffic flow and improving traffic safety. Beyond facilitating road safety via visual cues to drivers, the frictional properties of pavement marking surfaces are a pivotal element in safeguarding roadway safety. However, the friction characteristics of pavement marking surfaces have not been sufficiently investigated. Additionally, the integration of glass beads or other particles with pavement markings to enhance reflectivity and retroreflectivity complicates the study of their friction properties compared to bare pavements. To tackle these problems, this research utilizes the British pendulum tester (BPT), the circular track meter (CTM), the dynamic friction tester (DFT), and the three-wheel polishing device (TWPD) to evaluate the friction performance of various pavement markings. Eighteen specimen groups, comprising six types of markings (i.e., waterborne paint, preformed tape, epoxy paint, polyurea paint, MMA paint, and thermoplastics) with various glass beads and particles, were investigated to assess their impact on dry and wet friction, mean profile depth (MPD), and durability. The outcomes of this study serve as valuable resources for advancing safety measures and providing insights into emerging traffic management technologies.</p><p dir="ltr">Currently, there is an absence of established standards or methods for assessing and evaluating the friction characteristics of road markings. This lack of standardization has a pronounced impact on vulnerable road users-motorcyclists, bicyclists, and pedestrians-due to the potential for inadequate friction from road markings. To address the problem, this study has developed five friction levels based on the wet British pendulum number (BPN). Leveraging international standards and practical considerations, a tentative BPN range is advocated for crosswalks, symbols, and letters to enhance the safety of pedestrians and other susceptible road users.</p><p dir="ltr">Friction metrics, like MPD and friction number (FN), have been central to enhancing quality assurance and control (QA/QC) processes in chip seals. These metrics evaluate chip seal performance by examining problems such as aggregate shedding or significant bleeding, potentially leading to lower friction values or surface textures. However, instead of leading to slippery conditions, the loss of aggregate-particularly as a consequence of snow-plow operations-may result in the formation of uneven surface textures. The relationship between increased MPD or FN and enhanced chip seal quality is complex and not easily defined. This study introduces a groundbreaking method utilizing machine learning techniques, designed to improve the QC procedure for chip seals. A hybrid anomaly detection approach was applied to a dataset consisting of 183,794 MPD measurements, each representing the average mean segment depth (MSD) over 20-meter segments, gathered from real-world chip seal projects throughout the six districts managed by INDOT. A two-phase QC process, specifically tailored for chip seal quality assessment, has been developed. Validation analysis performed on four chip seal projects shows a strong concordance between field inspection, friction measurements, and the results predicted using the introduced approach. The developed method sets a foundational chip seal QC procedure, augmenting efficiency in acceptance processes and overall safety through data-driven techniques, while reducing the practitioners' time on site.</p><p dir="ltr">Surpassing the constraints of traditional approaches, this paper develops a series of scientific methodologies for evaluating friction on pavement and pavement marking surfaces through extensive in-field and laboratory experiments. Additionally, it establishes rational and efficient quality control procedures for chip-seal applications. The methodologies and conclusions presented in this paper can assist engineers in Departments of Transportation (DOTs) with ensuring the safety of all stakeholders, including road users, engineers, and construction practitioners. Furthermore, they offer valuable insights for the timely execution of road maintenance activities.</p>