هندسة تكنولوجيا المعلومات
Permanent URI for this collection
Browse
Recent Submissions
Now showing 1 - 3 of 3
- ItemModeling Adverse Events Detection as Time Series Using LSTM and Labeled Data Fusion(Al-Quds Univeersity, 2026-05-27) Rasha Zaki Mustafa Assaf; رشا زكي مصطفى عسافAdverse events (AEs) , characterized as undesirable and unintentional outcomes of medical treatments, pose significant challenges in healthcare. Accurate prediction of these events is crucial for enhancing patient safety, optimizing resource allocation, and improving overall health- care outcomes. Traditional methods, such as statistical analyses and early machine learning techniques, often fail to capture complex, nonlinear relationships of medical data. This dissertation explores the application of advanced machine learning models, particularly Long Short- Term Memory (LSTM ) and transformers, to predict adverse events and outcome using Food and Drug Administration (FDA ) Adverse Event Reporting System ( FAERS), a database containing information on drug safety, and Medical Entities in Digital Records Annotated ( Med- DRA) datasets. The dissertation focuses on integrating these comprehensive datasets, followed by data cleaning and pre-processing to ensure data accuracy and reliability. Latent Dirichlet Al- location( LDA) and clustering, a statistical method for topic modeling was used to address the complexity of classifying over 50,000 adverse event categories, reducing the number of classes to 10. The study further investigates the efficacy of sequence-to-sequence models, such as the Transformer, in predicting adverse events. Results indicate that Transformer-based models out- perform traditional machine learning algorithms in predicting adverse events, demonstrating improved accuracy and robustness. The sequence-to-sequence models achieved an accuracy of 92.81%, with a precision of 93.22%, a recall of 91.89%, and an F1 score of 92.55%. The integration of self-attention mechanisms enhances these models by allowing them to capture complex patterns and relationships within the data—insights that traditional approaches often overlook s. We implemented a Graph Neural Network ( GNN) and evaluated precision and re- call across multiple categories, yielding low overall F1 scores of 0.36 for term type classification and 0.42 for FAERS reaction classification.
- ItemTowards Automated Arabic Synonyms Extraction(Al-Quds University, 2025-01-11) Eman Abed Al-Kareem Mousa Naser; ايمان عبد الكريم موسى نصرSynonyms extraction gains special attention as synonyms are essential in improving Natural Language Processing (NLP) application performance. The Lexical Substitution (LS) task is utilized for Synonym extraction, which generates a set of equivalent substitutions (i.e., synonyms) to the target word or phrase in a sentence that saves the sentence's meaning. This task can enhance writing, language understanding, and NLP models and address ambiguity. Recently, LS has attracted much attention in many languages. Despite the richness of Arabic vocabulary, limited research has been performed on the LS task due to the lack of annotated data. To bridge this gap, we present the first Arabic LS benchmark dataset, AraLexSubD for benchmarking LS pipelines. AraLexSubD is manually built by eight native Arabic speakers and linguists (six linguist annotators, a doctor, and an economist) who annotate the 630 sentences. AraLexSubD covers three domains: general, finance, and medical. It encompasses 2476 substitution candidates ranked according to their semantic relatedness. We also present an Arabic LS pipeline, AraLexSubPro, which offers different techniques for generating, selecting, and ranking substitutions. To make a thorough comparison, AraLexSubPro uses four different methods as baselines to generate substitute candidates for the target words: a synonym dictionary-based approach using Arabic Word Net (AWN), a pre-trained language model-based approach (AraBERT), AraBERT dropout (partial masking), and a hybrid approach between AraBERT and AWN. The results showed that the hybrid approach achieved the best results compared to the other approaches. The generated substitutions are filtered and then ranked based on six high-quality features to compare thoroughly: word similarity, word frequency, BERT prediction order (BERT probability), BERT-based language model (Loss), BERT similarity, and the BERTscore. The substitutions are then reranked based on our AraLexSubPro ranker. Additionally, an error analysis of the experiment is reported. To evaluate the AraLexSubPro pipeline, we use our first benchmark dataset for the Arabic LS task AraLexSubD dataset, which can automatically evaluate the Arabic LS systems. To our knowledge, this is the first study on Arabic lexical substitution. The results were encouraging and fundamental for Arabic LS research. To speed up research on this field, we have put the AraLexSubD data on GitHub at the following link: https://github.com/karajah2024/Arabic-Lexical-Substitution.git
- ItemDiscovering Gene Associations Across Diseases Using a Knowledge-based Machine Learning Approach(Al-Quds university, 2024-11-04) Emma Mamdouh Jeries Qumsiyeh; ايما ممدوح جريس قمصيةComplex diseases such as diabetes, Alzheimer's, and cancer are influenced by a combination of genetic, lifestyle, and environmental factors that do not follow straightforward inheritance patterns. Biological systems are immensely complex and heterogeneous. To resolve the enigmas surrounding these systems, extensive research provides huge amounts of biological data. In this thesis and in our first study, a novel approach called GediNET was developed to integrate prior biological knowledge into disease-associated gene groups. GediNET employs a Grouping, Scoring, and Modeling (G-S-M) approach to identify top-performing gene groups, which are then used to train a machine-learning model. Following the data exploration and preprocessing steps, various classification models were built with 100-fold Monte Carlo Cross-Validation, and the performance of these models was evaluated. By applying Disease-Disease Association (DDA) based machine learning, GediNET uncovered new relationships between diseases, improving diagnosis, prognosis, and treatment approaches. In the second study, GediNETPro, an advanced version of GediNET, was developed. This version utilizes Cross-Validation (CV) information and clustering techniques, such as K-means, to identify patterns of disease group associations. GediNETPro provides visualization tools, like heatmaps and in-depth analysis of disease group clusters, offering insights for developing effective diagnostic interventions. The third study leveraged molecular-level data to develop effective methods for predicting Disease-Disease Associations (DDAs). A statistical technique was developed by employing the G-S-M-P model of GediNETPro to compute semantic similarity metrics between diseases. The semantic approach detects representative diseases within clusters and establishes a semantic relationship between the disease under investigation and other diseases. The studies presented in this thesis contribute to understanding disease complexity, uncovering disease associations, and identifying potential biomarkers and drug targets