Thesis in CSE

Browse

Filter results by typing the first few letters
Now showing 1 - 9 of 9
  • Thumbnail Image
    Item
    An efficient multicast routing protocol to minimize multipoint relays in MANET.
    (CUET, 21-May-2024) Hassan, Md. Zahid
    Reducing control packets, especially in proactive routing protocols, needed to establish
  • Thumbnail Image
    Item
    Assessing health related economic benefits from reduced particulate matter in air using BenMAP-CE
    (CUET, 4-Mar-2024) Kabir, Maisha
    The residents of Chattogram City Corporation (CCC) are grappling with serious health risks due to air pollution, especially during the dry period. The Institute for Health Metrics and Evaluation (IHME) has documented a significant number of premature deaths attributed to polluted air. While the government has implemented policies to curb air pollution, their impact often falls short due to a lack of comprehensive benefit modeling for decision-making. There is a lack of localized studies on air pollution and its health impacts, particularly in Chattogram, with gaps in understanding the correlation between air pollution, heavy metals, mortality, and economic burden, highlighting the urgent need for investigations using tools like BenMAP-CE to assess the efficacy of pollution reduction measures. This study bridges the gap by utilizing the BenMAP-CE tool to assess health and economic benefits resulting from airborne particulate matter reduction. The study centers around the development of BenMAP-CE, utilizing data on particulate matter (PM) in the air, population statistics, air pollution hazards as well as health hazards due to road dust, and economic data. PM data has been gathered from Landsat 8 and selected monitoring sites in the CCC area. Trace metals in road dust were identified using acid digestion and atomic absorption spectrophotometry.
  • Thumbnail Image
    Item
    Detecting financial fraud using Rule-Based Techniques.
    (CUET, 13-Feb-2024) Islam, Saiful
    Financial fraud is a growing problem that poses a significant threat to the banking industry,
  • Thumbnail Image
    Item
    Ergonomic analysis of seats of human powered vehicles by digital human modeling for better ride comfort
    (CUET, 3-May-2024) Hai, Tasmia Binte
    Rickshaws are essential for affordable and accessible transportation, particularly in densely populated urban areas. It is important to ensure the comfort of the rickshaw driver as it directly affects their health, well-being, and job satisfaction. It also influences customer satisfaction and safety. This study focuses on modifying the rickshaw driver's seat to enhance comfort while ensuring ergonomic principles are met. As proper cushioning, support, vibration absorption, and adjustability are key factors, this study involves measurements of rickshaw frames, CAD design of seat structures, ergonomic analysis using CATIA V5, and experimental vibration analysis.
  • Thumbnail Image
    Item
    Identification of cyberbullying Bangla Linguistics Texts using deep learning and transformer based approaches.
    (CUET, 24-Jul-2024) Saifullah, Md Khalid
    In today's digital era, social media platforms such as Facebook, Twitter, and YouTube play crucial roles in facilitating idea expression and interpersonal connections. However, alongside increased connectivity, these platforms have inadvertently facilitated negative behaviors, notably cyberbullying. While extensive research has delved into cyberbullying in high-resource languages like English, there remains a significant dearth of resources for low-resource languages such as Bengali, Arabic, Tamil, and others, particularly concerning language modeling. This study aims to bridge this gap by developing a cyberbullying text identification system, named BullyFilterNeT, tailored specifically for social media texts, with Bengali serving as a test case. The intelligent BullyFilterNeT system effectively tackles challenges associated with Out-of-Vocabulary (OOV) words inherent in non-contextual embeddings and addresses the limitations of context-aware feature representations. To provide a comprehensive analysis, three non-contextual embedding models—GloVe, FastText, and Word2Vec—are developed for feature extraction in Bengali. These embedding models are integrated into classification models employing both statistical methods (SVM, SGD, Libsvm) and deep learning architectures (CNN, VDCNN, LSTM, GRU). Furthermore, the study utilizes six transformer-based language models; mBERT, bELECTRA, IndicBERT, XML-RoBERTa, DistilBERT, and BanglaBERT to overcome shortcomings observed in earlier models. Notably, the BanglaBERT-based BullyFilterNeT achieves the highest accuracy of 88.04% in our test set, demonstrating its efficacy in identifying cyberbullying text in the Bengali language.
  • Thumbnail Image
    Item
  • Thumbnail Image
    Item
    Solving Blockchain Trilemma Using Off-Chain Storage Protocol
    (CUET, 4-Sep-2023) Reno, Saha
    Trilemma in blockchain refers to the infamous problem of simultaneously not delivering the
  • Thumbnail Image
    Item
    Text classification in a resource-constrained language using Deep Learning Techniques.
    (CUET, 23-May-2024) Hossain, Md. Rajib
    The exponential growth of unstructured textual data on the World Wide Web,
  • Thumbnail Image
    Item
    Trash Classification Using Deep Neural Network.
    (CUET, 20-Dec-2023) Das, Dhrubajyoti