Phd

Permanent URI for this community

https://escholar.umt.edu.pk/handle/123456789/4714

Browse

Browsing Phd by Issue Date

Filter results by year or month
Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Enhanced Diabetes Forecasting Model using EHR
    (UMT, Lahore, 2024) Iqra Naveed
    Diabetes mellitus Type 2 is being called a modern preventable pandemic. Investigating the massive EHR data to predict diabetes is the most significant challenge for the researchers. This thesis addressed this issue by investigating the gaps in the literature, highlighting the un-explored area, and proposing deep learning models. To attain this objective, current thesis explored variety of deep learning models that utilized the EHR data and contrasted with machine learning models to predict diabetes. Implemented models indicated deep learning models had greater potential to exploit longitudinal EHR data. To extend the research we investigated the predictive potential of features and explored the model performances of diabetic predictors, risk factors of diabetes, training data density and visit-history. This work exhibited that fasting blood sugar (FBS), hemoglobin (A1c), and BMI were the key predictors of diabetes.
  • Thumbnail Image
    Item
    Design of a Non-imaging Based Concentrated Photovoltaic System for Uniform Irradiance
    (UMT, Lahore, 2024) Waseem Iqbal
    Energy needs have been increased with the global advancement and industrial uprising. Electrical energy utilization shared a huge amount of energy to residential and industrial load. Traditional energy resources are expensive and pollutant, producing greenhouse gasses, which is the major environmental concern. Solar energy utilization is a cost-effective, sustainable, and green energy solution to meet the ongoing energy demand. Concentrator photovoltaic (CPV) systems are developed for energy conversion by providing high efficiency using multi-junction solar cells. The CPV system has been given preference over the photovoltaic due to its high efficiency. In a CPV system, most of the solar cell area has been replaced with an optical concentrator. Various parabolic trough based CPV systems have been presented where a concentration of <300× is achieved so far.
  • Thumbnail Image
    Item
    POWER QUALITY IMPROVEMENT WITH HYBRID POWER FILTERS USING SOFT COMPUTING TECHNIQUES
    (UMT, Lahore, 2024) Ayesha Ali
    The proliferation of electronic device usage in every field is deteriorating the system's Power Quality, causing both energy and economic losses. The motivation of this work rises from the scarcity of harmonic mitigation techniques to improve the power quality in local industries in Pakistan. Three harmonic elimination topologies are Passive Filters, Active Filters and Hybrid Filters. Hybrid filters bring the best of both passive and active filters together. The research work intends to implement a Shunt Hybrid Active Power Filter (SHAPF) with improvised control techniques to improve power quality indices. The optimized technique like an Artificial Neural Network (ANN), Gated Recurrent Unit, and Long Short-Term Memory network are implemented on MATLAB/Simulink to optimize the SHAPF performance. The novelty of the work is the implementation of deep learning methods to address problems with power quality.
  • Thumbnail Image
    Item
  • Thumbnail Image
    Item
    AI based NOMA for Heterogeneous Networks
    (UMT, Lahore, 2025) Syed Muhammad Hamedoon
    As wireless networks evolve to accommodate increasing user density, effective user clustering techniques are essential for optimizing resource allocation in Non-Orthogonal Multiple Access (NOMA) systems. Thus, as high-speed and quick data access becomes increasingly available, the need for a sophisticated and enhanced wireless network begins to emerge. NOMA-based heterogeneous networks are essential for the Internet of Things (IoT) era as they combine multiple wireless technologies and devices to enhance coverage, capacity, and user quality of service. In heterogeneous IoT networks with many users, exhaustively searching for optimal user pairs or groups becomes computationally infeasible. As user density increases, clustering becomes a combinatorial problem that demands scalable, low-complexity solutions. This research indicates various user partitioning algorithms that enhance network performance, spectral efficiency, and user fairness, particularly in dense scenarios with diverse channel conditions. We address the challenges of user clustering and power allocation in multi-carrier NOMA systems, emphasizing the importance of energy efficiency for IoT devices. A novel user clustering approach based on partial brute force search (P-BFS) is proposed, significantly reducing complexity while improving throughput. Additionally, we explore a Reconfigurable Intelligent Surface (RIS)-assisted NOMA framework that optimizes power allocation and phase shifts through advanced optimization techniques, including deep learning and reinforcement learning. RIS-assisted NOMA systems designed for IoT networks make better use of spectrum, which saves power and keeps connections between many users stable. This system differs from conventional wireless communication systems. This research uses AI-based methods to find the best balance between sum rate and energy efficiency, finding the best RIS phase shifts and power distribution. The results demonstrate substantial improvements in sum rate and energy efficiency, highlighting the potential of intelligent clustering methods for future 5G and beyond networks.