International Journal of Computer Engineering and Data Science (IJCEDS)

Genetic Algorithm-Based University Course Timetabling: A Practical Optimization Framework with Graphical Interface

2026-06-26T14:51:15+00:00

This article investigates the university course timetabling problem, a challenging combinatorial optimization problem encountered by educational institutions when allocating courses to available time slots and classrooms while satisfying numerous institutional requirements. A mathematical formulation is proposed distinguishing between hard constraints (strictly enforced) and soft constraints (penalised to improve timetable quality). Owing to the NP-hard nature of the problem, exact optimization techniques become computationally prohibitive for large-scale instances. Consequently, a genetic algorithm is developed with constraint-aware operators and a repair mechanism to efficiently explore the search space and generate high-quality feasible timetables. The proposed approach is evaluated on two realistic datasets involving 15 and 69 courses, using 10 independent runs to ensure statistical robustness. Results show that the algorithm achieves (100%) hard constraint satisfaction, an average room occupancy of (78%), with only minimal soft constraint violations (1 capacity overbooking and 2 room-type mismatches out of 15 courses), and generates feasible timetables in 45 seconds—compared to 4 hours of manual effort. A graphical user interface equipped with performance indicators is introduced to facilitate result interpretation and support administrative decision-making. These findings demonstrate that genetic algorithms provide an effective practical solution for university timetabling, with potential for deployment in medium-sized institutions.

Software Evolution Prediction Using Machine Learning Algorithms

2026-04-01T23:35:55+00:00

Software evolution represents the most time-consuming phase in the software development life cycle (SDLC) after a software release. Further, it covers an increasingly significant role in modern software development practices (e.g., Agile, DevOps, CI/CD) and web development (e.g., React, Next.js, etc.), where teams contribute more effort in improving and maintaining existing systems than building new ones. These evolutionary actions such as modifying and adding features, incorporating contributions, and fixing issues generate large software project data that demonstrate the dynamics of software development. Software evolution in machine learning (ML) techniques comprises the unending adaptation of models, data, and pipelines to continue performance under changing domains. In this paper, we are predicting the software evolution by analyzing features like Repository name, Repository link, Commits, Issues, Pull Requests, Stars, Forks, Total Contributors, Top Contributors, Code lines cover with Simple Linear Regression, Multiple Linear Regression, Polynomial Regression, and Support Vector Regression (SVR), are methods of ML. Thus, it is based on the results obtained a comparative study is conducted to evaluate and analyze model accuracy.

The Development of Digital Transcription Device Through ESP32 Integration: An Assistive Communication Tool for the Hearing Impaired

2026-02-26T00:11:51+00:00

Hearing impairment is a significant public health concern that continues to affect populations worldwide. The objective of this study is to provide hearing-impaired individuals with a real-time communication system that provides an additional communication support option, integrates artificial intelligence to ensure accurate communication, and addresses Sustainable Development Goal 10: reducing inequalities. The study aims to align with Qatar National Vision 2030 and current hearing impairment statistics in Qatar. This study employed a quantitative experimental research design to develop a Digital Transcription Device for the hearing-impaired as an assistive communication tool, integrating the ESP32. The study's results support the device’s effectiveness and accuracy, with a rapid response time of 3.56-9.17 seconds for segmented conversations. The device achieved 100\% accuracy with word counts ranging from 5 to 15. The effective distance of the Digital Transcription Device was found to achieve 100% accuracy at distances from 1 meter to 5 meters. Based on the findings, the Digital Transcription Device successfully enables accurate communication between hearing-impaired individuals by accurately recording the average time for words to display, the accuracy of the words displayed, and the maximum effective distance of the device, with minimal discrepancies.

TFRC-Based Selective Retransmission for Free-Viewpoint Video Streaming

2026-02-17T17:38:43+00:00

Free-viewpoint video may become the next big step in media technology that allows users to change the displayed viewpoint and synthesize custom views of a dynamic scene from a user-controlled perspective. The user-specific views are generated from two or more color and depth camera stream pairs that must be successfully delivered to the clients according to their continuously changing perspectives. Besides the view synthesis distortions, the higher error rate and limited bandwidth can cause quality degradation. We proposed an adaptive retransmission-based error recovery scheme allowing retransmissions according to the network congestion state and the allowed latency. The selective retransmission (TbSR) scheme for free-viewpoint video is based on the TCP Friendly Rate Control utilized as a bandwidth estimation algorithm. The TbSR approach was evaluated using the NS-2 network simulator and VSRS reference view synthesis tool in order to prove the efficiency of the proposed method.

Neuro Symbolic Automated Program Repair: A Systematic Review of LLM-Based and Symbolic Techniques

2026-03-09T16:50:01+00:00

Automated program repair (APR) involves creating patches for software defects with minimal human intervention. Classical template-based and search-based methods are not scalable, whereas large language models (LLMs) provide good generalization but are affected by hallucinations and have weak formal guarantees. Neuro-symbolic program repair (NSPR) integrates LLMs with static analysis, SMT solving, and symbolic techniques to trade off coverage, correctness, and interpretability. We performed a systematic review of APR and NSPR systems using the PRISMA 2020 guidelines. Predefined queries in IEEE Xplore, ACM Digital Library, Scopus, Web of Science, arXiv, and Google Scholar were used to search between January 2012 and January 2026. Architectures, benchmarks, outcomes, and deployments. Two reviewers screened the records against established predetermined eligibility criteria. A total of 70 pieces of academic primary empirical research on APR/NSPR and eight (8) pieces of industrial/production deployment reports were included, and 78 papers were ultimately included in the qualitative synthesis. Our results indicate that, compared to neural or symbolic systems, NSPR systems also tend to report higher correct repair rates and lower hallucination rates. It can be deployed in CI/CD pipelines, security patches, and large-scale open-source repositories. Nevertheless, standard benchmarks and transparent assessments may still be required to perform powerful comparisons and ensure reproducibility.

Digital Archive Classification Performance Analysis Using a Decision Tree Based on TF-IDF Features

2026-02-21T14:01:20+00:00

The management of digital archives at PT LNS Indonesia faces inefficiencies due to the absence of an automatic classification system, forcing employees to manually inspect file names. This leads to slow document retrieval processes and a high risk of misfiled archives. The aim of this study is to design and implement a digital archive classification system based on the Decision Tree algorithm to automate archive management. The research method follows the Knowledge Discovery in Databases (KDD) framework, starting from the selection of file names from four divisions Human Resources (HR), Finance, Information Technology (IT), and Engineering followed by text preprocessing and feature extraction using TF-IDF with a combination of word n-grams and character n-grams. The Decision Tree model is developed and evaluated using 5- Fold Cross-Validation. Experimental results show that the model achieves 91% accuracy, macro-precision of 0.92, macro-recall of 0.90, and a macro F1-score of 0.91. In conclusion, the implemented system successfully automates the grouping of thousands of files into folders according to division and document type, thereby significantly improving efficiency and accuracy in digital archive management within the company.

Data-Driven Success Prediction of Android Mobile Applications on the Google Play Store: A Systematic Literature Review (SLR)

2026-02-14T11:42:34+00:00

Every day number of mobile apps are added or removed from the Google Play Store platform depending upon app’s popularity in market place. Several attributes of an app, attributes that are either internal to or external to it, can perform a substantial role in deciding whether the app will be successful or not. In this paper, the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) are used for presenting a systematic literature review for success prediction for Android Mobile Apps before they are launched on the Play Store market. With the help of this SLR, developers and researchers can check existing models that can be used for predicting the success of an app. Apart from presenting existing approaches; the author has tried to assist in building a new success prediction model by identifying a few features/factors that can make a mobile soft-ware/application a success or failure. This paper formulates five Research Questions (RQs) and precedes the entire PRISMA process to find the answers to them. The answers to the RQs will help to devise a new success prediction model for mobile apps in the future.

QuakeApp: A Comprehensive Earthquake Detection and Monitoring System

2025-11-23T21:08:45+00:00

Traditional earthquake monitoring networks are limited by their sparse coverage and high deployment costs, leaving many at-risk areas without timely alerts. This work addresses whether smartphone accelerometers can be reliably used for real-time earthquake detection and how crowdsourced data can enhance the spatial density of seismic networks. QuakeApp was developed as a distributed sensing system, combining a mobile app for event detection, a scalable backend for data aggregation, and a dashboard for visualization. Previous research in similar crowdsourced systems has reported detection accuracies exceeding 90% under laboratory conditions, but QuakeApp’s results remain to be validated. The main contributions of this work are: (1) a distributed earthquake detection algorithm for smartphones, (2) a backend architecture for real-time crowdsourced aggregation, and (3) a quantitative assessment of citizen-based seismic sensing.

Sustainable Heatsink Optimization for Multi-Component PC enclosure

2025-11-07T20:29:49+00:00

Overheating is a significant challenge for the widespread use of desktop personal computers (PCs). This thermal issue frequently results in operational failures and poses safety risks. While current thermal optimization often relies on integrating a single heatsink (such as on the Central Processing Unit or CPU), research indicates this approach is insufficient for effectively lowering the machine's overall maximum temperature (T_max). To address this critical gap, this study introduces a novel passive multi-heatsink thermal solution. Utilizing Computational Fluid Dynamics (CFD) simulation, the research analyzes the strategic placement and performance of multiple generic heatsinks on various major heat-generating components inside the PC enclosure. The simulation successfully determined the optimum number and configuration of these components, verifying that distributing multiple passive heatsinks is a viable and effective method for significantly reducing the device's maximum operating temperature T_max.

Evaluating and Optimizing CNN–Transformer Architectures for Musculoskeletal Disease Classification

2025-10-06T22:04:22+00:00

This study examines the impact of dataset dimensionality on deep learning performance in musculoskeletal disease detection, focusing on osteoporosis and rheumatoid arthritis. Using over 200,000 annotated X-ray, DXA, and MRI images, the performance of Vision Transformer (ViT), ConvNeXt, and Swin Transformer models was systematically evaluated in terms of scalability, robustness, and multi-modal integration. Results demonstrate that increasing dataset scale significantly enhances model generalization, with Swin Transformer achieving the best performance (AUC = 0.94, p < 0.001). These findings underscore the critical role of self-attention mechanisms and model scaling strategies in medical image classification, providing new benchmarks for dataset requirements and guiding the development of more reliable AI-driven diagnostic systems. Furthermore, the study emphasizes the necessity of large, diverse datasets to mitigate overfitting and improve real-world applicability. It also highlights the potential of hybrid architectures for integrating multi-source medical data. Overall, this research contributes to advancing explainable and scalable AI solutions for musculoskeletal imaging in clinical practice.

Recyclitix: Waste Classification with CNN - Mobile Application

2025-09-29T16:37:37+00:00

Recylitix is an innovative mobile application that improves waste sorting efficiency through AI-powered image classification and contextual guidance. Leveraging computer vision and deep learning models using TensorFlow Lite, Recyclitix enables users to accurately identify waste types and receive localized recycling recommendations. This intelligent sorting mechanism reduces classification errors, optimizing recycling processes and minimizing the environmental impact of poorly sorted waste. The platform is built on a modern architecture that integrates a Spring Boot backend with a native Android application. Communication between components is facilitated by Retrofit for efficient API interaction. By combining robust machine learning with a user-centric mobile interface, Recylitix bridges the gap between sustainable practices and everyday behavior. It enables individuals, municipalities and waste management players to adopt smarter, more responsible recycling habits.

AI-VoiceTherapy: An Automated Platform for Voice Rehabilitation Using Artificial Intelligence

2025-09-29T16:38:29+00:00

AI-VoiceTherapy is a mobile platform that leverages artificial intelligence to democratize access to speech therapy. The system uses OpenAI's Whisper model to automatically detect and analyze speech disorders from voice recordings, including stuttering, dysphasia, dysarthria, and apraxia. Based on this analysis, the platform generates personalized therapy exercises tailored to the specific disorder and its severity. The three-tier architecture comprises an Android mobile application, a Spring Boot REST API, and a MySQL database. Key functionalities include automated speech analysis, personalized therapy generation, comprehensive progress tracking, and professional integration with speech-language pathologists. This innovation addresses geographical, economic, and resource barriers to traditional speech therapy, offering an accessible and scalable solution for millions affected by speech disorders worldwide.

A GIS-Based Geoportal for Land Management and Societal Acceptability in Mining: Case Study of MANAGEM Group, Morocco

2025-09-26T10:36:51+00:00

The complexity of land management coupled with societal considerations in Morocco's mining sector grows over time. Challenges such as land security, traceability of mining footprints, and social acceptability in sensitive territories necessitate integrated digital solutions. To allow societal considerations to be included in the land management for mining sites, this paper develops a bimodal solution. First, a web geoportal based on Geographic Information System (GIS) named MineMaps for centralised land data visualisation and management was developed. Then, a dynamic digital tool for structured societal actions’ planning aligned with Corporate Social Responsibility (CSR) and Environmental, Social and Governance (ESG) standards was elaborated. The case study is the Managem Group, a key player in the mining industry in Morocco. The geoportal enables interactive visualisation of mining sites and digitises land transactions, while the societal actions tool organises actions, stakeholder data, and impact indicators. Results demonstrate enhanced land traceability, reduced legal risks, and improved community engagement, fostering inclusive and sustainable territorial governance.

Mooditor: An AI-Powered Mobile Assistant for Real-Time, Emotion-Aware Mental-Health Support

2025-06-23T13:45:16+00:00

Mooditor is a pioneering mobile and web-based application designed to enhance mental health monitoring through artificial intelligence. By integrating real-time emotion detection via facial expression analysis and a Rasa-powered chatbot for therapeutic interactions, Mooditor provides a multi-modal approach to mental well-being. The system leverages computer vision and natural language processing (NLP) to assess psychological states, offering continuous monitoring and personalized support. Comprehensive tools, including mood tracking, statistical analysis, and conversation history, enable users and healthcare professionals to track emotional trends effectively. Our evaluation demonstrates exceptional performance, with the emotion detection model achieving a macro average precision, recall, and F1-score of 0.9998 across 953 instances. Mooditor’s modular architecture supports future enhancements, such as advanced emotion detection algorithms and integration with professional mental health services. This work addresses critical challenges in mental health accessibility and early intervention, contributing to the advancement of digital mental health care.

Keratoconus Classification Using Multimodal Imaging Strategy

2025-06-20T17:11:59+00:00

Data fusion improves the accuracy and robustness of diagnostic models by combining different types of information. This study presents a multimodal framework for keratoconus classification. It uses numeric and textual features from Pentacam reports, extracted with OCR. These are combined with corneal topographic images processed by a dual-branch deep neural network. The method was tested on 2,924 labeled Pentacam scans. Of these, 1,900 were used for training and 1,024 for testing. Scans were labeled as normal, suspicious, or keratoconus. Results show that combining image and text features improves classification. Deep learning accuracy rose from 96.78% to 98.34%. SVM improved from 93.35% to 95.60%. LDA increased from 92.85% to 94.80%, and KNN from 90.50% to 93.94%. These gains, up to 1.56% for deep learning and 3.44% for KNN, show the value of multimodal data for more accurate keratoconus diagnosis.

FitnityAI: Personalized Fitness Goal Tracking Assistant with AI

2025-05-15T11:21:42+00:00

With growing interest in health and wellness, there is a rising demand for intelligent tools that support personalized fitness routines. That’s what inspired FitnityAI, a fresh approach to mobile fitness tracking that uses generative AI to deliver customized guidance tailored to each user’s needs and progress. The application integrates key technologies, including the Gemini AI engine, a robust Spring Boot backend, and a user-friendly Android interface, to build a dynamic and adaptive fitness experience. Its standout feature is an AI-powered conversational assistant capable of interpreting user goals, activity patterns, and preferences to provide actionable, real-time fitness recommendations. FitnityAI was developed to support users in building sustainable fitness habits and to raise the bar for how mobile apps can use large language models to transform personal health management.

CityEcoScout: A Platform for Exploring Sustainable Locations Worldwide

2025-05-04T11:04:22+00:00

As people everywhere are trying to be more environmentally conscious, there is an increasing public demand for practical tools that help make sustainable decisions in cities and towns. That's what inspired CityEcoScout a new take on mobile platforms, using AI to inform users about environmentally focused spots in local areas and further afield. The application integrates several technologies, including Google Maps, Street View, and Places APIs, with the revolutionary Gemini AI engine to create something incredibly useful. The killer feature of this app is designed to provide easy-to-digest sustainability information and allows virtual tours of green destinations before arrival. CityEcoScout was developed to assist locals and travelers with making environmental choices, setting a new standard for how location services can enhance sustainability in daily lives.

Fashion Recommendation Systems: From Single Items to Complete Outfits

2025-04-23T16:29:02+00:00

Fashion recommendation systems have evolved beyond traditional recommender systems to address the unique challenges of fashion retail and e-commerce. This paper presents a comprehensive categorization of these fashion recommendation systems, grouping them into four fundamental approaches: personalization-based, compatibility-based, context-based, and special applications. We examine how personalization-based approaches leverage user preferences, while compatibility-based methods address fashion coordination through visual and semantic matching. The paper also explores the progression from single-item recommendations to complete outfit generation, alongside the integration of contextual factors like climate and occasions. Additionally, special applications such as body-shape awareness and sustainable fashion demonstrate the expanding scope of the field. Through this categorization, the paper provides a structured framework for understanding current approaches and identifying promising directions for future research, offering valuable insights for both researchers and practitioners in fashion recommendation systems.

A Multimodal Approach to Breast-Lesion Classification Using Ultrasound and Patient Metadata

2025-04-24T10:20:23+00:00

The diagnosis and prognosis of breast cancer have been greatly improved by incorporating machine learning methods, especially through medical imaging analysis as well as clinical information. In this study, the potential of deep learning models for breast lesion prognosis was explored by integrating imaging features with clinical data to enhance predictive accuracy. Clinical data were analyzed using multilayer perceptron (MLP) classifiers, XGBoost, and Random Forest, while several convolutional neural network (CNN) architectures, such as ResNet optimized with Adam, DenseNet with stochastic gradient descent (SGD), and EfficientNet with RMSprop, were evaluated. The integration of imaging-based features with clinical data was found to significantly improve model performance, enabling more accurate risk stratification and the development of individualized treatment strategies. The highest validation accuracy and area under the curve (AUC) were achieved by the most effective models, highlighting the advantages of a multimodal approach. Although the study was conducted on a relatively small dataset and faced challenges such as missing data, the results suggest that these methods hold considerable promise for implementation in clinical practice.

Healthcare Professional's Lifelong Learning Automation by Adapting Pedagogical Currents and Bloom's Taxonomy to Artificial Intelligence

2024-12-05T18:50:58+00:00

Background: This study focuses on the integration of artificial intelligence (AI), pedagogical techniques and Bloom's taxonomy in health sciences education. AI plays a key role in this field, changing educational paradigms through personalized learning experiences. Methodology: The study examines how AI enables personalized educational progression based on individual needs, promoting continuous lifelong learning. It examines the potential of AI to provide rapid feedback on tasks and assessments, improving conceptual understanding. In addition, AI helps trainers discover learning trends through data analysis, and creates dynamic learning environments. Results: Research shows that AI-based education systems boost students' grasp of complicated subjects, problem-solving ability, and writing capabilities. Furthermore, AI's flexible capabilities enhance educational inclusion by tailoring learning approaches to various individual problems. Discussion: The findings highlight AI's transformative impact on health sciences education, stressing the transition from traditional models to adaptive, learner-centered approaches. AI's ability to accommodate different learning styles and facilitate continual skill development demonstrates its promise to transform professional education in the health sciences. Conclusion: Integrating AI into health sciences education not only improves learning outcomes but also fosters a culture of lifelong learning among students and practitioners. As AI advances, its integration with pedagogical frameworks such as Bloom's taxonomy opens up new possibilities for improving educational procedures and preparing future healthcare practitioners for dynamic professional challenges.