UTAR Institutional Repository (Universiti Tunku Abdul Rahman)
Not a member yet
6132 research outputs found
Sort by
Overcoming the barriers to the adoption of green building in Malaysia
Get PDFThe Malaysian construction industry is a major contributor to energy consumption and greenhouse gas emissions, highlighting the importance of the sustainable solution, green building. Despite the existence of frameworks and policies such as the Green Building Index (GBI), adoption remains limited. This study investigates the barriers hindering implementation, evaluates the effectiveness of current green building policies, and proposes practical solutions to enhance uptake. A quantitative approach was employed through a survey questionnaire distributed to construction stakeholders, generating 60 valid responses analysed using SPSS Version 27. The findings reveal that high initial costs and lack of awareness are the most critical barriers. Although policies are generally clear in design, their effectiveness is hampered by weak enforcement and the limited accessibility of financial incentives. Respondents strongly supported the financial measures, such as tax rebates, complemented by nonfinancial strategies including professional training and stricter regulations. In conclusion, both financial and non-financial barriers are essential, and an integrated approach that combines financial, regulatory, educational, and technological measures is recommended to overcome the barriers and achieve Malaysia’s sustainability goals. Keywords: Green building, barriers, effectiveness, green building policies, solutions Subject Area: HD9715-9717.5 Construction Industr
Deep learning-based item classification for retail automation
Get PDFThis project focuses on developing a deep learning-based system for retail item classification.
The project aims to improve the efficiency and recognition accuracy in retail operations. By
leveraging the Convolutional Neural Networks (CNNs) and computer vision techniques such
as object detection and image segmentation. The project involved creating a comprehensive
dataset of retail items, which was preprocessed using data augmentation techniques to enhance
model generalization. The CNN model was optimized for both accuracy and speed,
incorporating regularization techniques such as dropout and batch normalization. Real-time
processing was achieved through the integration of object detection algorithms like YOLO and
image segmentation techniques. The final system was deployed and tested in a simulated retail
environment, where its performance was evaluated using metrics such as accuracy, precision,
recall and F1-score. This project also addresses the limitations of current methods and provides
a scalable solution for modern retail automation. Furthermore, this project contributes to the
field of retail automation by offering a scalable and adaptable solution for real-time item
classification. Additionally, it provides a practical framework for deploying deep learningbased classification systems in real-world retail settings, enhancing operational efficiency and
setting new standards for accuracy
Optimizing ChatGPT’s image analysis
Get PDFArtificial intelligence like ChatGPT has become a powerful tool in image analysis, but
its performance often declines when dealing with low resolution, reduced color depth,
blur, or noise. This project addresses these challenges through three key objectives
which is ChatGPT capability evaluation, ChatGPT preprocessing optimization, and
image quality restore module implementation. Together, these steps aim to establish
clear performance limits for ChatGPT, extend its capability through restoration
techniques, and design a robust pipeline for real-world applications.
Firstly, ChatGPT capability evaluation was conducted to determine ChatGPT’s
thresholds for reliable image analysis. Systematic testing revealed that a minimum
resolution of 512px with 24-bit RGB color depth provided the most consistent balance
between accuracy and efficiency. Performance dropped sharply when resolution was
reduced to 256px or lower, while higher distortion levels of blur and noise above 15%
significantly impaired accuracy. Further analysis showed that bar and line charts were
more vulnerable to distortion than pie charts, highlighting differences in sensitivity
across visualization types. These experiments established precise thresholds for
resolution, color depth, blur, and noise, providing a baseline for effective and consistent
analysis.
Furthermore, the model preprocessing optimization and image quality restoration
methods were explored to optimize good inputs and restore degraded inputs. For images
already meeting the thresholds, optimization was performed to reduce computational
load. The optimization is using an HSV-based background removal technique, where a
saturation threshold of 15 effectively reduced file size while maintaining accuracy. For
degraded inputs, an image restoration module was developed. Comparative testing
demonstrated that PixelCut AI consistently outperformed DeepImage AI in deblurring,
while a trained DnCNN model exceeded morphology-based approaches in denoising,
particularly under high noise levels. These findings confirmed that advanced restoration
techniques can extend ChatGPT’s capacity to analyse images that would otherwise fall
below acceptable quality levels.
Moreover, to optimize or restore the image, the implementation was realized through
an integrated image processing pipeline designed to balance efficiency with reliability.
The pipeline begins with quality evaluation to assess resolution, color depth, blur, and noise. Good-quality images are optimized to reduce computational load, while
degraded inputs undergo restoration through resizing, deblurring, or denoising until
they meet the minimum thresholds. A validation stage then ensures all processed
images satisfy the required standards before being analysed by ChatGPT. This structure
allows the system to optimize clear images while reliably enhancing poor-quality
inputs, ensuring consistent results across varied conditions.
The project culminated in the development of the ChatGPT AI Vision Assistant,
implemented in Streamlit, which supports both text and image queries while integrating
the image processing pipeline and a replot function for dynamic chart visualization.
This system enables users to test quality thresholds, experience optimized analysis and
observe the benefits of restoration methods. Overall, the project defines precise
thresholds for resolution, color depth, blur, and noise, enhances degraded images with
optimized and trained restoration models, and implements a complete pipeline that
balances efficiency with reliability. Together, these outcomes deliver a robust
framework that significantly strengthens ChatGPT’s reliability in real-world image
analysis tasks
Marker assessment, molecular characterisation, and phylogeny of Peninsular Malaysian Begonia spp. under Section Jackia
Get PDFBegonia L. encompasses over 2,000 species categorised into 70 sections. Peninsular Malaysia harbours numerous endemic Begonia species, many of which belong to Section Jackia, and are at risk of extinction. Most of these species are poorly identified and documented, with limited genetic and phylogenetic data available. This study employed nuclear (ITS) and chloroplast DNA markers (ndhA intron, ndhF–rpl32, and rpl32–trnL) as recommended by previous Begonia studies. These markers, either in individual or in concatenated form, were used to assess species delimitation and phylogeny of five Begonia samples from Peninsular Malaysia, alongside species from GenBank and earlier studies. The concatenated dataset (ITS+ndhA intron+ndhF–rpl32+rpl32–trnL) successfully delineated species but was unable to resolve the phylogenetic relationships between B. nurii and B. foxworthyi. Species delimitation analyses using ASAP, ABGD, PTP and GMYC revealed at least 14 species of Begonia in the dataset, although the classification of specimens BGNM1 and BGNG1 varied depending on the algorithm. This was further supported by genetic distance data, which recorded an interspecific genetic divergence of 0.51–2.16%. At least two species are potentially novel, while the taxonomic status of others (i.e., B. cf. nurii and B. cf. foxworthyi) remains uncertain due to missing or incorrect genetic data of type specimens. The study provided a preliminary phylogeny of Begonia within Section Jackia and highlights the rich diversity of Begonia in Peninsular Malaysia. Furthermore, the incorporation of floral morphological and genetic data of a wild-collected B. rajah confirmed its identity, establishing that it is extant in the wild after nearly a century of presumed extinction. The performance of each genetic marker (i.e., ease of amplification and sequencing, phylogenetic informativeness, etc.) was assessed, laying valuable groundwork for advancing taxonomic research optimised for the fast-evolving Begonia species of Southeast Asia
Fraud detection using machine learning in e-commerce
Get PDFThe fast growth of e-commerce has resulted in a rise in fraudulent activities, posing significant challenges to the security and trust of online transactions. Traditional fraud detection methods often fall short in effectively identifying complex fraud patterns due to issues like data imbalance, misclassification of costly errors, and the evolving nature of fraud tactics. This research proposes a machine learning-based approach to improve fraud detection performance in e-commerce platforms. Resampling techniques like SMOTE, oversampling and under-sampling are applied to address class imbalance issue. The study aims to reduce false negatives and enhance the detection of rare fraudulent transactions. Ensemble models such as Random Forest, AdaBoost, and XGBoost, will be employed to capture complex patterns and improve model performance. A systematic model evaluation was conducted using metrics such as accuracy, F1-score, MCC, precision, recall and AUC to ensure robust performance. Experimental results showed that Random Forest combined with oversampling achieved the best trade-off between precision and recall, reducing false negatives while maintaining high overall accuracy. Robustness was further validated through testing on both synthetic datasets and the Kaggle dataset, confirming the model’s adaptability and reliability. Finally, the best-performing model was integrated into a Power BI dashboard, enabling real-time monitoring of fraud detection results and visualization of emerging fraud trends. This integration supports decision-making by providing stakeholders with timely insights. The study contributes to the development of adaptive fraud detection systems capable of mitigating financial risks and maintaining customer trust in the e-commerce sector
Kindergarten management system
Get PDFThe Kindergarten Management System (KMS) project aims to improve administrative efficiency in kindergartens by addressing the limitations of manual record-keeping, ineffective communication, and lack of digital payment options. Most kindergartens still use old-fashioned methods, which often lead to delays, mistakes, and things running less smoothly day to day. This project seeks to enhance an existing KMS by incorporating better communication tools, stronger security measures, and online payment integration.
The new system will add messaging, email alerts, and virtual meetings to make it easier for teachers, parents, and school staff to stay connected and communicate smoothly. Multimedia support for announcements and Q&A sections will provide clearer and more engaging communication. To ensure data security, the project will implement Multi-Factor Authentication (MFA), password encryption, CAPTCHA verification, and automated backups, protecting sensitive school information from unauthorized access. A safe online payment system makes collecting fees easier and faster, so there's less manual work and payments come in on time.
This project builds upon existing KMS platforms by addressing their weaknesses and incorporating advanced features. The result will be a more efficient, secure, and user-friendly system that improves communication, strengthens security, and simplifies administrative tasks in kindergartens
Personalized nutrition and obesity intervention system
Get PDFObesity is a condition characterized by an excess accumulation of body fat which is a major health problem prevalent in millions of people across the world and there is a consistent rise in prevalence in both adults as well as youths. In this project we propose the creation of a Personalized Nutrition and Obesity Intervention System which is investigated by providing the user with personalized diet options based on personal profiles in Mobile Application. This app is made with a primary focus on personalized nutrition, in contrast to many other apps that consider nutrition as a secondary feature. Some of its features include a menu that can be adjusted depending on the end user, features that allow users to monitor their health and features that help to promote engagement and continued use of the application. With the help of an integrated smart nutrition calculator in the mobile application, users may input personal information like age, weight, gender, activity level, and health goals to receive personalized recommendations for calorie intake. The dietary planner ensures that consumers may modify their meal plans as their health needs change by providing flexible customization based on dietary choices, allergies, and real-time feedback. There is also a diet and health option on the software where the user can record the food they take and other health aspects such as weight and Body mass index (BMI). Charts and graphs are used to show users daily activity and goals progress. Like in most mobile applications, motivational features such as daily check-ins, challenges, and point-based reward systems present in the app to enhance the users’ experience. The basic features of such applications are returning points to the activity in hope of receiving vouchers or discounts to enhance retention. In addition, multilingual support ensures that the users with different languages and culture can easily access the application thus allowing it to be used by a global population of users. Java was the main programming language used in Android Studio. To ensure its reliability and scalability other tools such as Git version control and XAMPP control panel were used for backend services. The project addresses several of the fundamental limitations of current mobile health solutions, such as the difficulty of sustaining long-term good eating habits and the lack of customization in obesity intervention tools. Through the aid of the app, the target is to fill a gap in the market by offering specialized nutrition which will empower and encourage user to achieve their desired wellness and health in nutrients
Skin analysis and recommendation system
Get PDFPerceived opacity and uneven performance of AI skin analysis tools drive users toward unguided, trial-and-error use of skincare products, amplifying ingredient conflicts and delaying improvement. To address this issue, this project develops a skin analysis and recommendation system that delivers a trusted skin analysis model alongside a content-based recommendation method. Firstly, the system provides real-time face detection using the Android library and produces results from the deployed model accompanied by accuracy metrics. This enables use across diverse environments and ensures reliable results through guidance on capture quality and low-confidence flags. To achieve robust skin analysis process, Several AI models, including Classic CNN, EfficientNetB0, ResNet50, and GPT assistant–based models, were trained and tested on datasets for acne severity level and skin type classification. Among these, the YOLOv8 model demonstrated superior performance, achieving testing accuracies of 76.2% for acne severity and 64.0% for skin type, outperforming all other models. The integration of GPT-4o introduces a GPT assistant–based approach that complements traditional deep learning by enhancing interpretability and flexibility in prediction tasks through concise rationales and uncertainty cues. Secondly, the system generates skincare products recommendations based on either the current analysis results or user-specified conditions through content-based filtering. The content-based filtering process applied predefined features rules, whereby acne severity level and skin type govern the selection of recommended product attributes. Users can then view detailed information about these products, with the system generating ingredient-based justifications to support informed decision-making. To enhance personalization, the system has been integrating LLM models, allowing users to select either the Gemini API or the OpenAI API, which provide additional insights and recommendations. Thirdly, selected products can finally be added to a personalized skincare routine, where the system performs AI analysis to evaluate overall compatibility of the routine with the user’s skin condition. Fourthly, the system further aggregates skin analysis results to track progress trends for acne severity and skin type in overall, daily and monthly basis, where the system will perform descriptive analysis on progress trend to provide user with useful insight on their skin condition. This project successfully developed multiple module which are skin analysis, product recommendation, skincare routine management, and skin progress tracking to provide a seamless and reliable skin analysis and skincare recommendation experience while ensuring a user-centric and privacy-conscious mobile design
GreenDefender: Predictive plant health and care system
Get PDFThis project explores the intersection of artificial intelligence (AI) and mobile technology to provide home gardeners with a personalized, user-friendly solution for maintaining decorative plant health. Existing applications such as Plantix, Agrobase, PlantVillage and Blossom largely focus on agricultural crops, and offer limited disease coverage, minimal personalization, and inadequate real-time support for ornamental plants. GreenDefender addresses these gaps by integrating a multi-stage AI pipeline, an interactive recommendation system, and a cloud-enabled user experience into a single mobile platform. The proposed application was trained using two publicly benchmarked datasets: the Leaf Detection Dataset from Kaggle, containing 1130 annotated images for verifying the presence of leaves, and a combined disease dataset comprising the Rose Lead Disease Dataset, Flower Leaf Diseases Dataset New and Update, Tomato Leaf Disease Dataset, and Pumpkin Leaf Diseases Dataset from Mendeley Data, providing 9683 images across six disease classes: Fresh Leaf, Mosaic, Powdery Mildew, Downy Mildew, Black Spot, and Rust. Both datasets were split using an 80:10:10 ratio for training, validation, and testing. The application employs a YOLOv8-based leaf detection model as a gatekeeper to verify the presence of leaves before disease classification, achieving results with an mAP50 of 0.961 (training), 0.748 (validation), and 0.681 (test).As for disease classification, 3 deep learning architectures – Attention U-Net, Standard U-Net, and ResNet50 – were trained and compared. Attention U-Net achieved the highest accuracy, with 97.94% test accuracy (random splitting) and 96.42% (stratified splitting). Beyond detection and classification, GreenDefender provides interactive care plan generation powered by Gemini Artificial Intelligence (AI). Users answer context-specific questions to receive highly tailored recommendations. The application also supports diagnosis history tracking, profile management, and a community discussion module to encourage knowledge sharing among users
Environmental, social and governance (ESG) considerations: building occupations' perspectives
Get PDFThe real estate sector significantly contributes to climate change, accounting for approximately nearly one-third of carbon emissions and over a third of energy consumption. As individuals spent approximately 90% of their time indoors,
this raised important concerns regarding health, well-being, and quality of life. While numerous studies explored ESG practices in relation to financial performance, limited research has focused on ESG practices from the perspective of building occupants. Therefore, this study aims to uncover the
ESG considerations from building occupants’ perspective in Malaysia. The literature review identified eighteen (18) important ESG practices, which served as the basis to examine twenty-three (23) ESG criteria, grouped into nine (9)
environmental, eight (8) social, and six (6) governance aspects. A quantitative approach was employed in this study, where an online questionnaire was distributed to building occupants of high-rise buildings in Klang Valley. A total
of 148 responses were collected and analysed using the Cronbach's Alpha Reliability Test, Arithmetic Mean, Mann-Whitney U Test, Kruskal-Wallis Test, and Spearman’s Correlation Test. The Arithmetic Mean results revealed that
safety and security, along with health and well-being, were perceived as the most important ESG practices. Occupants placed the highest value on criteria such as the installation of security systems and the adoption of energy-efficient
technologies. However, the adoption of renewable energy and green certifications remained low, with only security systems and natural lighting being commonly implemented. The Mann-Whitney U test indicated that tenants, married individuals, and highly educated occupants significantly prioritised
affordability, cost reduction, and resource efficiency. Meanwhile, the Kruskal Wallis test identified significant differences across age, property age, education level, and geographical location, highlighting the varying ESG preferences among occupants with different demographic and residential backgrounds. A strong correlation was identified using Spearman’s Correlation test between “enhancing water efficiency” and “practicing water conservation.” The findings
underscore the importance of integrating occupant perspectives to support sustainable development goals and to encourage responsible property management within Malaysia's urban residential sector.
Keywords: Environmental; Social; Governance; ESG practices; Building occupants
Subject Area: HT101-395 Urban groups. The city. Urban sociolog