MSA Repository "MSAR"

MSAR University's Digital Repository is a documentation and digitization of all university outcomes that are of effective value in the scientific and academic community and reflects the university's image, work, and effective contribution to society Through MSAR Digital Repository, the university managed to collect, store, archive and publish digital content - including documents, audio files, images and data sets - all in a safe place. MSAR is one of the strongest University Digital Repositories in Egypt and documented in the DSPACE community with its latest versions.

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 5 of 6

Recent Submissions

  • Thumbnail Image
    Item type: Item ,
    STEAD-PE: A Structured Static-Dynamic PE Framework for Arabic Medical Response Generation
    (Association for Computing Machinery, 2026-10-11) Ahmed Alansary; Molham Mamdouh; Mohammad Moftah; Ali Hamdi; Khaled Shaban
    Medical response generation in low-resource and linguistically complex languages such as Arabic requires precise, context-aware advice, making Prompt Engineering a critical mechanism for controlling large language model behavior and improving output quality. The demand for scalable and reliable Arabic medical support systems has increased interest in prompt-based controllability, particularly where model fine-tuning and large annotated datasets are impractical. However, existing LLMs often show limited clinical relevance and semantic alignment when prompted with unstructured, patient-authored Arabic medical queries. In this work, we propose STructurEd stAtic-Dynamic PE (STEAD-PE), a prompt-only framework for Arabic medical response generation. STEAD-PE operates without model fine-tuning or parameter updates and leverages expert-curated Arabic medical posts paired with clinically validated treatment plans. The framework introduces case-specific structured metadata Domain, Severity, and Diagnosis—dynamically injected at inference time to enable clinically grounded and context-aware responses. We evaluate STEAD-PE across thirteen state-of-the-art LLMs, including Arabic-specialized, multilingual, and general-purpose models, using BERTScore for semantic similarity. Results show consistent and monotonic improvements in semantic and clinical relevance across all models, with STEAD-PE (Dynamic) achieving the highest BERTScore-measured semantic alignment in every case. Among evaluated models, LFM2-350M attains the best overall BERTScore (86.50%), while Qwen-32B improves from 80.89% (Zero-Shot) to 84.90% (Static) and 85.76% (Dynamic). These findings establish STEAD-PE as a robust, model-agnostic PE framework and highlight the importance of dynamically injected clinical context in reducing linguistic ambiguity and improving semantic fidelity in low-resource medical NLP settings.
  • Thumbnail Image
    Item type: Item ,
    Regime-dependent bitcoin volatility forecasting under alternative information sets: evidence from econometric, deep learning, and hybrid models
    (Springer International Publishing, 2026-10-26) Ahmed S. Wafi; Sherif El-Halaby; Sameh Abd-ElMaksoud Aboul-Dahab
    Bitcoin volatility is challenging to forecast because accuracy depends on model choice, information sets, and market regimes. Existing studies often compare models under inconsistent volatility targets, predictors, samples, or evaluation designs, leaving unclear whether forecast gains arise from model architecture, information content, or regime shifts. This study addresses this gap by proposing a unified rolling out-of-sample framework for daily Bitcoin volatility forecasting that compares GARCH-type models, standalone LSTMs, and state-augmented GARCH–LSTM hybrids under the same volatility target, nested information sets, one-step-ahead forecast horizon, and evaluation design, using balanced daily data from 1 February 2018 to 19 July 2025. Predictors are organized into technical (T), technical plus sentiment (TS), and technical, sentiment, and blockchain (TSB) groups to measure the marginal value of added information. Forecast accuracy is assessed mainly with QLIKE, plus RMSE, MAE, and Diebold–Mariano tests across normal and crisis regimes. Predictability is strongly regime-dependent. In normal markets, LSTM-T achieves the lowest loss, with TGARCH and Hybrid_GARCH_T statistically indistinguishable from it. In crises, hybrids minimize QLIKE, but the best specification differs: Hybrid_TGARCH_TS during COVID-19, Hybrid_EGARCH_TSB during the China mining ban, and Hybrid_APARCH_TS during the crypto winter. Hence, hybrid gains are regime-specific and depend on the interaction of volatility dynamics, information sets, and regime characteristics. For daily one-step-ahead Bitcoin forecasts, the evidence supports regime-sensitive modeling and cautions against broad claims of model superiority based on pooled samples.
  • Thumbnail Image
    Item type: Item ,
    Effect of Incorporating Perceptual Learning Into Plantar Hardness Discrimination on the Risk of Falling and Muscle Torque in Patients With Hemiplegia
    (Lippincott Williams and Wilkins Ltd., 2026-10-29) Lama Saad El-Din Mahmoud; Hanaa S. Mohamed; Marian M. Shafeek; Mohammed Saad El-Din Mahmud; Hanan Hosny M. Battesha
    Objectives: This study aims to evaluate the impact of perceptual training involving plantar hardness discrimination on fall risk and muscle torque in individuals with hemiplegia. Design: A randomized controlled trial was conducted in thirty-four stroke patients were randomly assigned to the study and control groups in equal numbers and assessed both before and after treatment. The outcome evaluations comprised fall risk assessment by the Biodex Balance System (BBS), the Isokinetic assessment of lower-limb muscle torque, and the Functional Independent Measures Scale (FIM). Each group was given a selected physical treatment program, while the study group also received perceptual learning with plantar hardness discrimination. Results: Post-treatment comparison revealed that the study group exhibited a significantly lower risk of falling than the control group (P <0.001, effect size [ES]=1.48). Similarly, knee extensors torque increased significantly in the study group relative to the control group (P <0.001, ES=1.19). Additionally, the study group showed a significant rise in FIM scores compared with the control group (P <0.001, ES=1.56). Conclusions: For stroke patients, incorporating perceptual learning into plantar hardness discrimination in the rehabilitation program is regarded as beneficial physical therapy neurorehabilitation for enhancing risk of falling, lower limb muscle torque and function.
  • Thumbnail Image
    Item type: Item ,
    Applying Dempster Shafer Theory and Causal Inference in Investigating Internet of Things Digital Forensic Challenge
    (Science Publications, 2026-10-10) Mohamed El-Dosuky; Sherif K. Ratib; Mostafa M. El-Gayar
    The rapid growth of IoT ecosystems has transformed digital forensics, creating challenges in analyzing distributed, heterogeneous, and uncertain evidence. Conventional forensic methods often struggle to handle ambiguous and conflicting IoT data, highlighting the need for robust frameworks that effectively manage uncertainty across multiple evidence sources. This research presents a hybrid forensic framework that integrates Dempster Shafer Theory (DST) for probabilistic uncertainty modeling with causal inference for deterministic reconstruction of event relationships. Together, these methods provide a comprehensive approach to evidence fusion and event attribution in complex IoT investigations. Using the DFRWS 2017 IoT Digital Forensic Challenge as a case study, the framework systematically processes and correlates diverse evidence sources, including Google Hangouts messages, Alexa voice logs, IoT sensor traffic, and wearable activity data. Results show that DST-based evidence fusion reduces the DST-based epistemic uncertainty (i.e., the proportion of uncommitted belief/ignorance) from 18% before evidence fusion to 0.2% after integrating all available evidence, while achieving a belief value of 97.4% for suspect identification. Complementary causal inference analysis reconstructs temporal and behavioral dependencies, improving the interpretability of probabilistic outcomes. The combined model enhances reliability, transparency, and reproducibility, offering a scientifically grounded methodology suited to modern, data-rich IoT forensic environments where traditional techniques fall short.
  • Thumbnail Image
    Item type: Item ,
    Explainable Hybrid Intelligence for Autonomous Cyber Defense: Adaptive Intrusion Detection and Real-time Firewall Enforcement Using Reinforcement Learning
    (Intelligent Network and Systems Society, 2026-08-12) Magi Hossameldin Mahfouz; Raghda Essam Ali; Ayat Mahmoud
    Conventional rule-based firewalls rely on static signatures, limiting adaptability to unseen attacks, while AI-enabled commercial tools remain proprietary and lack transparency. This paper presents Net-Knight, an explainable hybrid intelligence framework integrating supervised intrusion detection, unsupervised anomaly detection, reinforcement learning (RL)-based mitigation, and Linux nftables enforcement. A LightGBM classifier trained on 22 engineered NetFlow features from NF-UQ-NIDS-v2 performs intrusion detection, while an autoencoder ensemble supplies complementary anomaly evidence. These signals form a 28-dimensional state used by a MaskablePPO agent to select among five mitigation actions. The IDS achieves 96.82 accuracy and a 0.9681 macro F1-score across nine classes; the RL agent attains a 96.9 rule-match rate; and the integrated decision pipeline reaches 96.3 end-to-end decision accuracy at 8.7ms mean processing latency in the controlled evaluation. A proposed Hierarchical Confusion-Aware Disambiguation (HCAD) extension is evaluated separately through Monte Carlo sensitivity analysis; its projected gains are not treated as experimentally measured performance.