Artificial Intelligence

Permanent URI for this collection

https://hdl.handle.net/20.500.11888/20217

Browse

Browsing Artificial Intelligence by Title

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    A HYBRID DEEP LEARNING MODEL FOR FORECASTING PM2.5 AIR POLLUTANT CONCENTRATIONS
    (An-Najah National University, 2024-12-18) Massad, Asma
    Air quality forecasting is a crucial research field that aids scientists and policymakers in making informed decisions to combat air pollution. Among various pollutants, PM2.5 -particulate matter with a diameter smaller than 2.5 micrometers- poses significant health risks, as it can reach the lower respiratory tract and enter the bloodstream. Accurately forecasting PM2.5 levels is thus essential. Although machine learning-based spatiotemporal forecasting models have advanced, the pursuit for more accurate forecasts continues. The use of hybrid deep learning models for PM2.5 forecasting represents a promising and active area of research, as these models aim to capture complex spatiotemporal dependencies more effectively. We developed a Dynamic Graph Attention Network (DyGAT) to model spatial dependencies effectively. DyGAT leverages engineered edge features, including distance, wind speed, and wind direction, while using attention mechanisms to capture the dynamic nature of these dependencies. DyGAT was then combined with Informer, a Transformer for efficient time-series forecasting, to capture spatial and temporal patterns comprehensively, improving prediction accuracy. Our model was evaluated on a benchmark dataset from Beijing, with 420,768 records over four years. DyGAT-Informer outperformed a version without the DyGAT component and other baseline models. It achieved 50.43 for MAE, 79.9 for RMSE and 28.88% for SMAPE, compared to 51.44 for MAE, 80.83 for RMSE and 30.25% for SMAPE in the next best model. Additionally, we conducted a case study using a dataset from Nablus, Palestine, consisting of 2692 records per station over a two months period. We incorporated geospatial features about nearby pollution sources into the dataset. Due to the insufficient number of records in the Nablus dataset for training the Informer, it was replaced with a sequence-to-sequence Long Short-Term Memory (LSTM) model. DyGAT-LSTM, trained with additional geospatial features about nearby pollution sources, achieved a 2.08% reduction in MAE, 1.17% in RMSE, and 1.96% in SMAPE. This confirms the benefit of incorporating such data. Finally, despite the short distances between stations, DyGAT successfully captured spatial dependencies, where DyGAT-LSTM achieved a reduction of 3.13% in MAE, 1.48% in RMSE, and 3.67% in SMAPE when compared to the LSTM-only model.
  • Thumbnail Image
    Item
    COMPUTER VISION AND MACHINE LEARNING APPROACHES FOR INTERPRETATION OF BREAST HISTOPATHOLOGY
    (An-Najah National University, 2024-09-12) Odeh, Omar
    Breast cancer, along with other cancer types, represents a major health issue worldwide, especially in regions with limited medical resources, such as Palestine where it’s harder to diagnose and treat cancer patients. The traditional diagnostic methods which rely mainly on the manual examination of histology slides in the lab by experts are both time consuming and require high skills, which often leads to delays in diagnosis and treatment. This research aims to leverage Whole Slide Imaging (WSI) and advanced image processing techniques, including machine learning and deep learning, to develop a tool to support breast cancer diagnosis in Palestine. A collection of histopathology images from breast cancer cases, provided by An-Najah Hospital in Nablus, Palestine, was used in this study. An automated classification model was developed, capable of identifying normal, benign, in situ, and invasive breast cancer tissues at magnifications of x20 and x40. It was empowered using advanced preprocessing techniques to normalize the colors and sizes of the training data. The classification model resulted in a remarkable performance of InceptionV3 and Gradient Boosting models with accuracies of up to 98.7% and 98% respectively in different settings. The same algorithm was applied to a BACH challenge dataset, resulting in a high accuracy of 96.7%, surpassing the results of other studies. The study also covers whole slide image (WSI) analysis, emphasizing the complex approach required for the proper detection of malignancy and tumor size estimation using a customized segmentation approach which resulted in an 84% Jaccard index score. Additionally, it discusses novel approaches in localization of tumor beds and grading of treatment responses, aiming to provide a methodological basis for improving patient care outcomes in challenging environments like Palestine. Keywords: Breast Cancer Diagnosis, Automated Image Analysis, Whole Slide Imaging (WSI), Deep Learning in Pathology, Healthcare in Palestine.
  • Thumbnail Image
    Item
    IMPROVING ARABIC E-LEARNING USER EXPERIENCE THROUGH SENTIMENT ANALYSIS AND COLLABORATIVE FILTERING MODELS
    (An-Najah National University, 2025-07-26) Yamin, Aya Said
    This study works to develop an enhanced recommendation system (RS) for the purpose of improving the user experiences in e-learning environments by achieving the integration between sentiment analysis (SA) and collaborative filtering (CF). It uses a publicly available English dataset from Coursera and translates it into Arabic using AWS translation services to be appropriate for the target language context. It targets Arabic-language course reviews, taking into consideration handling data sparsity and linguistic complexity challenges. A refined multilingual Bidirectional Encoder Representations from Transformers (BERT) model was used to produce sentiment labels, and support vector machine (SVM) classification with a combination of Term Frequency–Inverse Document Frequency (TF-IDF) and FastText features achieved good performance. And then the user-item interaction matrix was enriched with the sentiment scores that resulted from SA to make recommendations using Alternating Least Squares (ALS) for active users and using K-means clustering based on profiles, followed by hybrid K-Nearest Neighbors (KNN) and TF-IDF similarity on Arabic course names for cold users. The evaluation of the system is done by making a comparison before and after adding the effect of sentiment separately for active and cold users. The system specifically targets users or learners who plan to study Arabic courses on online platforms. The findings emphasize that this integration of sentiment information reduces the limitations related to cold-start user problems and also enhances personalization. For example, the sentiment-aware model achieved a reduction in RMSE of nearly 70% for active users and a significant improvement in success rate (from 19.57% to 93.27%) and recall (from 18.68% to 91.24%) for cold users. These improvements lead to considering it a practical, useful solution for e-learning platforms.
  • Thumbnail Image
    Item
    MACHINE LEARNING ALGORITHMS FOR PREDICTING STUDENTS' ACADEMIC PERFORMANCE IN EDUCATIONAL DATA MINING
    (An-Najah National University, 2025-06-21) Salmiyah, Fatimah
    Student performance prediction has been one of the important works in educational data mining, due to the possibility of early detection, intervention, and informed decision-making in academics. The purpose of this study is to improve the accuracy of predicting student performance by using seven machine learning models—Decision Tree, Random Forest, Linear Regression, Neural Network, Support Vector Machines (SVM), Logistic Regression, Naive Bayes and five feature selection techniques : Particle Swarm Optimization (PSO), Lasso, Wrapper Method, SelectKBest and SelectPercentile. The research investigates how student outcomes are associated with such factors as previous educational experience, parental education, past educational failures, attendance, residence and participation in extra-curriculum activities. The results show that Linear Regression combined with SelectKBest achieved the highest accuracy of 93.5% . The performance of the model was further optimized by hyperparameter tuning (GridSearchCV) and k-fold cross-validation, which increased both prediction accuracy and model robustness. The results highlight the role of feature selection in maximizing model performance and offer practical guidance for higher education institutions interested in implementing predictive analytics to enhance student success.
  • Thumbnail Image
    Item
    NEBRAS: A RAG-BASED QUESTION ANSWERING SYSTEM FOR ISLAMIC AND LEGAL GUIDANCE
    (An-Najah National University, 2025-02-27) Al-Huwari, Samer
    Question answering (QA) systems are essential tools in natural language processing (NLP), designed to interpret user queries and generate relevant answers. These systems have evolved over time from rule-based models to advanced machine-learning-based approaches. The emergence of the transformers architecture and Large Language Models (LLMs) have set the stage for modern QA systems. LLMs have transformed QA by leveraging vast datasets to generate human-like responses across various domains and their ability to understand complex linguistic patterns. However, LLMs often generates plausible but incorrect answers particularly in specialized domains like law and religion where accuracy is critical. This phenomenon is known as “hallucination”. The risk of “hallucination” is increased when dealing with a complex language like Arabic. Arabic language, with its rich morphology, diverse dialects, and its dependency on diacritics, present significant challenges for LLMs primarily trained on Western languages. Fine-tuning LLMs for domain-specific tasks is time-intensive, and computationally-expensive, given their massive parameters size, demanding innovative approaches to mitigate the LLMs hallucination issue without extensive re-training. This thesis introduces Nebras, a generic multi-domain QA system leveraging a Retrieval-Augmented Generation (RAG) framework, LLM agents, and a hybrid retrieval approach. Nebras’s knowledge base can be dynamically extended by following simple guidelines and using its built-in mapping component, enabling it to adapt to any textual dataset. By employing an Agentic RAG pipeline, Nebras optimizes each processing stage using specialized agents. Furthermore, it utilizes pre-trained LLMs without fine-tuning, enhancing scalability and reducing computational costs. Experimentation results demonstrated Nebras’s performance in Arabic domain-specific QA. In the Islamic fatwa domain, it achieved a BERTScore-F1 of 70.94%, a METEOR of 13.49%, with 9 accepted fatwas compared to only 7 accepted from GPT-4o. In the university help-desk domain, Nebras achieved a BERTScore-F1 of 75.80%, METEOR of 40.20%, and BLEU of 9%, significantly outperforming the BLEU score of 2.3% from GPT-4o's. These results highlight Nebras's ability to enhance factual accuracy, confirming its potential as a scalable Arabic QA solution.