Journal of King Abdulaziz University: Computing and Information Technology Sciences

Human Activity Recognition with Smartphones Using Machine Learning

Wejdan S Alghamdi, Ameerah Alshahrani , Nehal Otaif, Najlaa Alqurashi — Mon, 30 Dec 2024 00:00:00 +0000

Human activity recognition is widely used now in many applications, such as smart homes, health care, and business as well as in a wide range of pattern recognition and human-computer interaction research. In this paper, we use the same sensors embedded in smartphones (Accelerometer and Gyroscope) to track and recognize human activities. We employ a machine-learning algorithm, which is the Support Vector Machine (SVM) to improve the performance of human activity recognition. The experimental results on the HAR Dataset from the UCI repository indicate that our approach is practical and achieves 96% accuracy.

Social Media Enabled Public Value Creation for A Saudi Arabian Municipality: A Critical Realism Paradigm Perspective

Dr. Turkey Althagafi, Abeer Alghamdi — Mon, 30 Dec 2024 00:00:00 +0000

Globally, government organizations including local government agencies have become progressively interested in using social media applications to open new venues of interactions with citizens. Due to the nature of social media applications in terms of users’ ability to generate content, a higher level of engagement is expected to take place for not only to deliver public services but also to design as well as render innovative public services. Despite the growth in literature on social media, there is still a limited understanding on what mechanisms should be employed to create public value by using social media applications. This study therefore aims to identify the causal mechanisms and other enabling other conditions that jointly explain public value creation using social media applications. To address this concern, we report on the development of a model to investigate public value creation using social media applications. The model has been empirically evaluated using a qualitative case study in a large Saudi Arabian Municipality from a critical realism perspective. The model and empirical evidence together contribute towards establishing a theoretical foundation for research into the impact of social media applications for public value creation. In addition, municipality managers can learn useful lessons drawing on our findings. The study also presents a methodological contribution to social media research by providing insights into the application of critical realism ontology and methodology for assessing public value creation through the use of social media applications.

A Multi-Label Code Comment Classifier using BERT

Zarah Shibli, Emad Albassam — Mon, 30 Dec 2024 00:00:00 +0000

Code comments play an essential role in software development by providing documentation, explanations, and clarifications for program logic and functionality. It is crucial to effectively classify code comments to improve software maintainability and collaboration in the face of a growing amount of code. Developers can easily identify and comprehend different code sections' purpose, behavior, and requirements by accurately classifying code comments. This paper presents a novel approach utilizing multi-label classification to enhance code comment classification in three programming languages: Python, Pharo, and Java. We employ BERT, a widely used language model, and achieve an F1 score of 0.64 through experimentation. Our proposed approach facilitates the understanding and managing code comments, making software development more efficient and productive. Additionally, our approach can be extended to other programming languages and serve as a foundation for further research in code comment classification.

Investigating Active Learning based on Dynamic Data Selection techniques for Image Classification

Salma Kammoun — Mon, 30 Dec 2024 00:00:00 +0000

This paper explores the efficacy of Active Learning (AL) techniques, specifically focusing on Dynamic Data Selection (DDS), for improving image classification tasks. AL is a machine learning paradigm that enables the automatic selection of the most informative data samples for annotation, thereby reducing the annotation burden and enhancing model performance. In this study, we investigate the integration of DDS techniques with AL strategies to iteratively select the most informative image samples for model training. We use a fine-tuned VGG16 as the underlying classification model due to their effectiveness in image analysis tasks. Our experimental evaluation involves comparing the performance of fine-tuned VGG16 trained with three AL-based DDS techniques on Arabic sign language dataset. We analyze various DDS strategies, including Random selection, Entropy-based selection, and margin selection to determine their impact on model accuracy and annotation efficiency. The results of our study demonstrate the effectiveness of margin selection method-based AL approach in improving the performance of recognition of 32 hand gestures for Arabic sign language (95.3 \%) while minimizing the annotation effort.

Recent Advances in Dysarthric Speech Recognition: Approaches and Datasets

Tahani Alrajhi, Mourad Ykhlef, Ahmed Alsanad — Mon, 30 Dec 2024 00:00:00 +0000

Dysarthria is a neuromotor speech disorder that results from physical disability and limits speech intelligibility. Dysarthric speakers can make use of speech recognition systems to help them communicate more effectively with others. This paper surveys the latest works conducted on dysarthric speech recognition that was carried out in a span of five years, specifically from 2018 until 2023. These works are categorized according to the approach that was followed to improve dysarthric speech recognition. The approaches include data augmentation, enhancement of dysarthric speech, speech and acoustic features, adaptation, and hybridization of multiple approaches.

Optimizing Federated Learning for Medical Image classification: A Comparative Study of Pre-Trained Models on Compressed X-ray Imager

Sawsan Alwadaie, Dr.Amani, Dr.samar, Dr. Lamiaa — Mon, 30 Dec 2024 00:00:00 +0000

Machine learning, particularly deep learning, has revolutionized a number of fields, including medical diagnostics. In this study, federated learning is employed to address privacy concerns and data access limitations inherent in medical imaging. A simulated FL environment was used to investigate the performance of five pre-trained neural network models: DENSENET121, RESNET18, VGG-NET11, GOOGLENET and INCEPTION-V3. It emphasizes the optimization of training duration as well as the application of lossy image compression techniques such as JPEG in order to improve communication efficiency. We conducted a comparative analysis of the models' performance before and after image compression by evaluating the Area Under the Receiver Operating Characteristic Curve and the training time. According to the results, image compression can maintain or improve model performance while affecting training time, underscoring the trade-offs between model accuracy and computational efficiency.

Tasaheel-v2: Development of Innovative Textual Analysis tool with Advanced Features

Hanen Hemdi, Fatmah Assiri — Mon, 30 Dec 2024 00:00:00 +0000

We introduce Tasaheel-v2, an automated tool specifically developed for Arabic Natural Language Processing (NLP)
and textual analysis tasks. This work is an extension to the first version, Tasaheel-v1, comprised of traditional NLP tasks
including stemming, segmentation, normalization, named entity recognition, and part of speech tagging. Furthermore, it
included cutting-edge analytic methods, such as specific emotion, polarity, linguistics, and domain-specific word tagging. In this
new innovative version, Tasaheel-v2, we introduce additional benefiting utilities designed to provide assistance for the Arabic
research community. We specifically integrate another Arabic-specific POS tagger, a sentiment analyzer, and English-to-Arabic
translation functions. We leverage the utilities provided in Tasaheel to develop a machine-learning model designed to identify
Arabic phishing emails and provide a thorough textual analysis to capture deceptive cues used to detect phishing linguistic
patterns. This tool contributes to the Arabic research domain by providing assistive NLP functions and textual analysis features
all in one tool