Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
55 results
Search Results
Item Restricted Machine Learning Systems for Unsupervised Time Series Anomaly Detection(Saudi Digital Library, 2025) Alnegheimish, Sarah; Veeramachaneni, KalyanModern assets – from launched satellites to electric vehicles – output dense, multivariate time series data that must be monitored for deviations from “normal” behavior. This monitoring task is referred to as time series anomaly detection. The current state of the industry still depends on fixed or heuristic thresholds that often drown operators in false alarms, and can miss the subtle, context-dependent faults that matter most. This thesis addresses unsupervised time series anomaly detection as an end-to-end problem, asking how we can learn, evaluate, and deploy models that judiciously flag anomalies while remaining intuitive to the end user. This thesis provides contributions in the form of both algorithms and systems. First, it introduces three models that enlarge the design space of unsupervised time series anomaly detection: TadGAN, which leverages adversarial reconstruction; AER, which unifies predictive and reconstructive objectives in a single hybrid score; and MixedLSTM, which explicitly incorporates interdependencies to improve anomaly detection in multivariate time series. We propose two range-based evaluation metrics that quantify detection quality over temporal intervals. Second, it presents our system Orion, which abstracts anomaly detection pipelines as directed acyclic graphs of reusable primitives, providing user-friendly APIs and enabling interactive visual inspection. Building on this infrastructure, OrionBench performs periodic, fully reproducible benchmarks, producing leaderboards that align research innovations with the needs of end users. Third, the thesis explores a new paradigm – foundation models for unsupervised time series anomaly detection – by formulating SigLLM, which employs large language models and time series foundation models for zero-shot anomaly detection via prompting and forecasting. This paradigm indicates a promising path to developing scalable models for anomaly detection. Finally, beyond evaluating our systems on publicly available datasets, we provide extensive experiments on two industrial case studies that demonstrate improved detection accuracy and practical usability of our system.25 0Item Restricted Machine Learning-based Detection Strategies for DDoS Attacks(Saudi Digital Library, 2025) Alshmlan, Abdullah Salem A; Songfeng, LuWith the rapid development of information technology, Distributed Denial-of-Service (DDoS) attacks have become a major threat to network security, posing severe challenges to the online services of enterprises and individuals. Traditional defense methods are often inefficient against complex, evolving attack patterns and fail to provide better detection and response. To address these limitations, this study focuses on developing and evaluating machine learning-based models for detecting Distributed Denial-of-Service (DDoS) attacks. A hybrid model combining lightweight Convolutional Neural Networks (CNNs) and Bidirectional Long Short-Term Memory (BiLSTM) networks is developed to leverage CNN’s spatial feature extraction and BiLSTM’s temporal dependency modeling. The CIDDS-001 dataset is used after rigorous preprocessing, including cleaning, feature selection, normalization, and sliding-window segmentation. Several architectures are trained and compared, including the proposed CNN-BiLSTM and an enhanced Self-Attention BiLSTM variant that dynamically emphasizes critical traffic patterns. Experimental evaluation using metrics such as accuracy, precision, recall, and F1-score demonstrates that the hybrid and attention-based models achieve superior performance and effectively reduce false alarm rates. Overall, the study provides a practical and adaptable approach for DDoS attack detection, enhancing the responsiveness and reliability of network defense systems. Future work will focus on extending this framework to larger and more diverse datasets to further improve its generalization in real-world scenarios.10 0Item Restricted Scientific Portfolio Optimization: A Risk-Adjusted Approach to Asset Allocation(Saudi Digital Library, 2025) Alawad, Naif; Neil, PhillipsThis dissertation evaluates the robustness of traditional, risk-based, and machine learning (ML) portfolio optimization methods under realistic market conditions. Classical Mean–Variance Optimization (MVO) is elegant in theory but fragile in practice due to estimation error and instability in crises. Risk-based approaches such as Risk Parity (RP) and Hierarchical Risk Parity (HRP) provide more resilient alternatives by allocating on volatility and correlation structures instead of unstable return forecasts. ML-enhanced MVO (ML-MVO), which substitutes predicted for historical returns, remains of uncertain value. A modular Python artefact was developed to compare these strategies using rolling five-year windows, monthly rebalancing, and strict walk-forward validation, complemented by an interactive dashboard interface. Performance was assessed through risk-adjusted metrics (Sharpe, Sortino, maximum drawdown, volatility) across both normal and crisis regimes, including the Global Financial Crisis (2008–2009) and the COVID-19 shock (2020). Sensitivity analysis with realistic weight constraints was also conducted to test robustness under practical implementation settings. Results show HRP consistently achieved the most robust risk-adjusted outcomes, outperforming MVO and ML-MVO in both full-sample and stressed settings. RP and equal weighting remained competitive baselines, while ML-MVO underperformed despite moderate predictive accuracy. Overall, the findings suggest ML contributes more effectively to restructuring optimization processes, as in HRP, than to direct return forecasting. The study also highlights inherent limitations of short-horizon ML forecasting and points to future research extending horizons, incorporating richer features, and exploring ML-enhanced risk estimation.6 0Item Restricted Graph Neural Networks for Drug Screening(Saudi Digital Library, 2025) Aqeeli, Noura Eissa; Panas, DagaDrug discovery is a lengthy and costly process that often involves small, noisy, and imbalanced datasets. In our study, we investigate the use of graph neural networks (GNNs) for predicting molecular homeostatic activity in neuronal cells through transfer learning. We evaluate Graph Convolutional Networks (GCNs) and Message Passing Neural Networks (MPNNs) with transfer learning, comparing their performance to Random Forest and non-transfer GNN baselines. To guide the selection of source datasets for pre-training, we implement a molecular latent representation similarity framework across nine MoleculeNet datasets. Additionally, we fine-tune a foundational molecular model on our target dataset. We evaluate the models using five-fold cross-validation, using the Area Under the Receiver Operating Characteristic curve (AUC-ROC) and the Area Under the Precision-Recall curve (AUC-PR) as metrics. Our results indicate that transferring knowledge from high-similarity source datasets outperforms the baseline models. Moreover, source-to-target transfer is more effective than fine-tuning the foundation model; however, the foundation model exhibits superior generalisation capabilities. Finally, we employ a selected set of models to rank an unlabelled molecular dataset. Our findings demonstrate that GNNs, combined with similarity-guided transfer learning, enhance performance in predicting bioactivity within low-data and imbalanced settings, highlighting the importance of carefully selecting source datasets to avoid negative transfer.9 0Item Restricted Multi-Omics Approaches to Explore Vancomycin Treatment Mechanism in Patients with Primary Sclerosing Cholangitis (PSC) - Inflammatory Bowel Disease (IBD)(Saudi Digital Library, 2025) AlOmar, Haneen; Acharjee, AnimeshIntroduction: Primary sclerosing cholangitis (PSC) is a comorbid condition associated with inflammatory bowel disease (PSC-IBD) that lacks effective treatments beyond liver transplantation. Although oral vancomycin (OV) has shown therapeutic promise, disease activity often returns after treatment withdrawal. This study aims to investigate the mechanisms of OV in PSC-IBD patients, supporting the development of more durable and targeted therapies. Method: Paired multi-omics data from 15 patients before and after OV treatment were analysed. The datasets included RNA-Seq, metatranscriptomics, bile acid metabolites, and 16S rRNA. After preprocessing, feature selection was performed using LASSO, ElasticNet, and Boruta-RF. Selected features were analysed in two complementary ways: first, intersected features that were identified by all models were assessed for their predictive robustness and integrated into correlation network graphs. Union features were subjected to pathway enrichment analysis to elucidate their biological significance. Results: The 3 models consistently selected a total of 13, 2, 4, and 3 intersected features simultaneously from RNA-Seq, metatranscriptomics, bile acid metabolites, and 16S rRNA, respectively. These features achieved predictive performance comparable to or superior to the full datasets. For example, intersected features outperformed the full dataset in metatranscriptomics, where Boruta-RF achieved a higher AUC (0.936 vs. 0.896), demonstrating the robustness and efficiency of selected features. Pathway enrichment analysis of union features in each omics revealed pathways related to mucosal healing, metabolism, and immune modulation. Correlation networks graphs demonstrated that OV-induced alterations in cross-omics before and after treatment. Conclusion: Based on paired data from only 15 patients, this study provided a comprehensive multi-omics perspective on OV’s impact in PSC-IBD patients and identified robust biomarkers. We also uncovered novel host–microbiome interactions not previously reported, highlighting potential targets for future therapies. While findings are promising, they require validation in larger, independent cohorts.10 0Item Restricted Predicting Osteoarthritis in Older Adults Using Literature-Based, Non-Invasive Risk Factors: A Cross-Sectional Analysis of ELSA Wave 9(Saudi Digital Library, 2025) Fnais, Tesneem; Yang, HuiOsteoarthritis (OA) is a prevalent joint disorder in older adults that is often diagnosed at a later stage, as clinical assessments typically rely on imaging and laboratory tests that are not readily accessible in all settings. This study aimed to develop and evaluate machine learning models that predict OA using non-invasive, self-reported features from Wave 9 of the English Longitudinal Study of Ageing (ELSA). A total of 4,723 participants aged 60 and above were included. An initial set of 32 features was selected based on existing literature and refined through a structured feature selection pipeline, resulting in a final set of 25 features, including joint pain and mobility limitations. Four supervised models -Logistic Regression, Random Forest, XGBoost, and CatBoost- were trained using a stratified train-test split and resampling to address class imbalance. The upsampled logistic regression model achieved the highest sensitivity (0.769) and strong overall performance (AUC = 0.755), while CatBoost showed the highest specificity (0.759) and an AUC of 0.747. A reduced logistic regression model using only the top 15 features retained similar accuracy and AUC. These findings demonstrate that OA can be predicted without imaging or biomarkers. The resulting models, particularly the logistic regression model, offer promise as cost-effective screening tools to support early identification and guide decisions about further clinical assessment. making them well-suited for primary care and digital health settings, especially where resources are limited.9 0Item Restricted Evaluating Machine Learning for Intrusion Detection in CAN Bus for in-Vehicle Security(Saudi Digital Library, 2025) Alfardus, Asma; Rawat, DandaThe past decade has seen a potential rise in the automobile industry accompanied by some serious challenges and threats. Increased demand for intelligent transportation system facilities has given a boom to the automotive industry. A safer and better experience is much sought from vehicles. It opens opportunities of including autonomous vehicles and Vehicle to Everything technologies in the automotive sector. Enabling vehicles to connect to various services exposes to compromise and misuse by the adversaries. There are numerous electronic devices in the modern vehicle which communicate with each other using multiple standard communication protocols. State-of-the-art vehicles are the assembly of complex mechanical devices with the sophisticated technology of electronic devices and connections to the external world. Controller Area Network (CAN) is one of the widely used protocols for in-vehicle communications. However, the lack of some fundamental security features such as encryption and authentication in CAN makes it vulnerable to security attacks. The backbone of connecting autonomous vehicles is CAN with limited bandwidth and exposure to unauthorized access. Various attacks compromise the confidentiality, integrity, and availability of vehicular data through intrusions which may endanger the physical safety of vehicles and passengers. These security shortcomings, therefore, lead to accidents and financial loss to the users of vehicles. To protect the in-vehicle electronic devices, researchers have proposed several security countermeasures. In this work, we discuss various security vulnerabilities and potential solutions to CAN’s. Further, a machine learning-based approach is also developed to devise an Intrusion Detection System for the CAN bus network. This study aims to explore the adaptability of the proposed intrusion detection system across diverse vehicular architectures and operational conditions. Furthermore, the findings contribute to advancing the state-ofthe-art in automotive cybersecurity, fostering safer and more resilient transportation ecosystems. Moreover, it investigates the scalability of the intrusion detection system to handle the increasing complexity and volume of data generated by modern vehicles.21 0Item Restricted Towards Industrially Adoptable Generation Invariant Reprocessable Polydicyclopentadiene Thermoset Plastics(Saudi Digital Library, 2025-05) Alfaraj, Yasmeen; Johnson, JeremiahThe industrial transition to sustainable polymer technologies necessitates novel end-of-life approaches for historically un-recyclable thermoset plastics. Polydicyclopentadiene (pDCPD), a high-performance thermoset known for its superior mechanical and thermal properties represents a compelling target for sustainability-oriented innovation due to its established industrial use, diverse manufacturing methods, historic challenges in reprocessing, and an increased interest from its relevant industries to recover valuable fillers and reinforcing materials from pDCPD carbon-fiber-reinforced polymers (CFRPs). Recent reports exhibit the ability to deconstruct pDCPD through a cleavable comonomer (CC) approach; however, we currently lack cost-effective strategies for scaling its deconstruction and recycling. This thesis addresses the fundamental barriers to industrial implementation of deconstructable pDCPD thermosets through a comprehensive, three-pronged approach that integrates data-driven molecular design, drop-in strategies for multigenerational recyclability, and cost-informed evaluation of CCs. In the first part of this work, a closed-loop experimental–computational platform is developed to predict glass transition temperatures (Tg) in deconstructable pDCPD networks incorporating bifunctional silyl ether (BSE) CCs and cleavable cross-linkers. Leveraging a curated dataset of 101 compositionally diverse pDCPD-based thermosets, machine learning model ensembling and strong regularization techniques are implemented to mitigate overfitting and quantify predictive uncertainty. Experimental validation of model predictions shows that the resulting models achieve accurate Tg predictions for variable CC and cleavable cross-linker loadings, novel CCs, and previously unseen related classes of strand cleaving cross-linkers. This chapter demonstrated the viability of predictive informatics in navigating the vast chemical and compositional space of deconstructable thermosets. The second segment presents a minimally chemically intensive, drop-in strategy for pDCPD recyclability. Using cleavable BSE comonomers and cross-linkers, networks with up to 20 wt% recycled oligomeric fragments are synthesized and evaluated. These materials exhibit thermomechanical properties and deconstructability that remain invariant across three generations of recycling. Furthermore, the incorporation of a cleavable cross-linker, dimethyl di-dicyclopentadiene silyl ether (DDMS), not only preserves but enhances bulk properties such as Tg in virgin and recycled samples, and addresses issues of oligomer incorporation in recycled samples as evidenced by gel fraction analysis. The ability to maintain and tune materials properties without post-processing or structural reformulation underscores the industrial potential of the drop-in CC approach for scalable, circular thermoset manufacturing. The final component of the thesis evaluates MeSi7, a seven-membered BSE CC, as a low-cost, synthetically accessible, and possibly scalable alternative to existing CCs. Thermodynamic polymerization parameters and CC performance under industrial thermoset cure conditions are assessed. We find that high-temperature cure conditions enable sufficient incorporation into the pDCPD network strands for deconstruction with as low as 5 mol% loading of MeSi7. These samples retain Tg values above 100 °C, with a moderate reduction relative to non-deconstructable analogues. Assessment of performance in industrial formulations also shows comparable deconstructability thresholds and modest impact on Tg. Importantly, MeSi7 is projected to cost less than 2% of iPrSi8 based on raw material pricing, offering a highly attractive economic profile for broader market applications. Together, these contributions deliver a framework for the rational design, performance prediction, and techno-economic evaluation of cleavable, recyclable thermosets through a convergence of data science, molecular design, and systems-level engineering considerations.9 0Item Restricted SEVERITY GRADING AND EARLY DETECTION OF ALZHEIMER’S DISEASE THROUGH TRANSFER LEARNING(Saudi Digital Library, 2025) Alqahtani, Saeed; Zohdy, MohamedAlzheimer’s disease (AD) is a neurological disorder that predominantly affects individuals aged 65 and older. It is one of the primary causes of dementia, and it contributes significantly and progressively to impairing and destroying brain cells. Recently, efforts to mitigate the impact of AD have focused with particular emphasis on early detection through computer aided diagnosis (CAD) tools. This study aims to develop deep learning models for the early detection and classification of AD cases into four categories: non-demented, moderate-demented, mild-demented, and very mild demented. Using Transfer Learning technique (TL), several models were implemented including AlexNet, ResNet-50, GoogleNet (InceptionV3), and SqueezeNet, by leveraging magnetic resonance images (MRI) and applying image augmentation techniques. A total of 12,800 images across the four classifications that were preprocessed to ensure balance and meet the specific requirements of each model. The dataset was split into 80% for training and 20% for testing. AlexNet achieved an average accuracy of 98.05%, GoogleNet (InceptionV3) reached 97.80%, ResNet-50 attained 91.11%, and SqueezeNet 86.37%. The use of transfer learning method addresses data limitations, allowing effective model training without the need for building from scratch, thereby enhancing the potential for early and accurate diagnosis of Alzheimer’s disease [1].17 0Item Restricted Improving Sleep Health with Deep Learning: Automated Classification of Sleep Stages and Detection of Sleep Disorders(Saudi Digital Library, 2024-07-07) Almutairi, Haifa; Datta, AmitavaSleep consumes roughly one-third of a person’s lifetime, and it is characterized by distinct stages within sleep cycle. The sequence of these stages at night provides insights into the quality of sleep. Poor sleep quality can have numerous consequences, including drowsiness, reduced concentration, and fatigue. Beyond sleep quality, an analysis of the sequence of sleep stages can uncover the presence of sleep disorders. This thesis aims to focus on three key research problems related to sleep. Firstly, it focuses on the classification of sleep stages using a combination of signals and deep learning models. Sleep stages are categorized into five distinct stages, namely Wake (W), non-rapid eye movement (NREM) stages comprising N1, N2, and N3, and rapid eye movement (REM) stage. Throughout the duration of sleep, individuals experience multiple cycles of sleep stages. Each cycle contains a standard allocation of each stage. An unbalanced distribution of the stages can indicate the presence of sleep disorders. Previous studies primarily classified sleep stages using a single channel of electroencephalography (EEG) signals. However, incorporating a combination of signals from electromyography (EMG) and electrooculogram (EOG) alongside EEG data provides additional features. These features extracted from muscle activity and eye movements during sleep, thereby enhancing classification accuracy. In this thesis, a robust model called SSNet is proposed to accurately classify sleep stages from a fusion of EEG, EMG, and EOG signals. This model combine convolutional neural networks (CNNs) and long short-term memory (LSTM) networks to extract the salient features from various physiological signals. The CNN architecture extracts spatial features from the input signals, while LSTM architecture captures the temporal features present in signals. This study has obtained encouraging outcomes in the classification of sleep stages through the fusion of physiological signals and deep learning techniques. Secondly, this thesis aim to detect obstructive sleep apnoea (OSA) from electrocardiography (ECG) signals using deep learning methods. Sleep disorder breathing (SDB) is categorized into three different types, which are OSA, central sleep apnoea, and mixed sleep apnoea. OSA is the most common form of SDB that is characterized by repeated interruptions in breathing during sleep, leading to fragmented sleep patterns and various health complications. Previous studies developed feature engineering methods and machine learning models for the detection of OSA. Feature engineering methods involve crafting relevant features to feed into machine learning models. However, feature engineering is time-consuming and requires domain expertise. In contrast, deep learning automatically extracts features from ECG signals for OSA detection, eliminating the need for manual feature engineering methods. In this thesis, three deep learning architectures are proposed, including standalone convolutional neural networks (CNN), CNN with long short-term memory (LSTM), and CNN with gated recurrent unit (GRU). Through rigorous experimentation and evaluation, the combination of CNN and LSTM architecture is the best-performing model for OSA detection. To further enhance the architecture’s performance, the hyperparameters of the CNN and LSTM models were tuned and tested over a large dataset to validate their effectiveness. The third research problem addressed in this thesis is detection of periodic leg movements (PLM) and SDB from NREM stage by using a combination of signals and deep learning models. PLM is characterized by involuntary leg movements during sleep. These movements can disrupt sleep and result in daytime sleepiness with reduced quality of life. Detecting PLM and SDB events during NREM stage allows for quantifying the severity of sleep disorders. Previous studies have focused on the development of signal-based models for detecting PLM or SDB. However, the models lacked the ability to distinguish these events within specific sleep stages. To address this problem, a novel deep learning architecture known as DeepSDBPLM is proposed. This architecture aims to detect PLM and SDB events during the NREM stage. This architecture incorporates novel input features called attention EMDRaw signals and utilizes a Residual Convolutional Neural Network (ResCNN) model. This thesis presents experimental results using publicly available datasets to evaluate the performance of the proposed deep learning models for classification of sleep stages, and detection of sleep disorders. The models were evaluated standard metrics. It includes accuracy, sensitivity, specificity, and F1 score. The empirical results establish the effectiveness of proposed approaches. The models can be a stepping stone towards more advanced techniques.25 0