UTAR Institutional Repository (Universiti Tunku Abdul Rahman)
Not a member yet
6132 research outputs found
Sort by
Developing an e-commerce website for enhancing customer sales
Get PDFThe rapid growth of e-commerce, particularly within the fashion industry, has
significantly transformed how businesses interact with customers. This project
addresses common challenges e-commerce platforms face, such as inadequate product
details, suboptimal mobile user experience, limited customer support accessibility, and
the lack of engaging features that cultivate customer confidence. These issues have
hindered e-commerce websites from maximising customer sales and enhancing user
satisfaction. To overcome these challenges, the project aims to create a responsive,
user-friendly web shop that integrates comprehensive product information, improves
mobile functionality, and facilitates customer support with AI product
recommendations between the AI agent and buyers. Additionally, the platform will
integrate gamification elements, like interactive quizzes and reward systems, to boost
consumer engagement and retention. The project will ensure structured development
using the Waterfall model within the Software Development Life Cycle (SDLC). The
target users include online shoppers, customer service agents, marketing teams, and
business managers. The project aims to provide personalised product suggestions,
enhance customer interaction, and ensure smooth device navigation. With the
integration of a dashboard displaying relevant statistics, the project will also aid
management by providing valuable insights into user behaviour and sales trends,
ultimately helping to optimise business strategies with AI sentiment analysis based on
user reviews in the admin dashboard. This project is significant as it aims to enhance
the e-commerce experience, increase customer satisfaction, and drive higher sales,
contributing to the growth and success of the digital marketplace
Efficient similarity index based task routing for LLM
Get PDFLarge Language Models (LLMs) represent a significant innovation within the field of
Generative AI (GenAI), yet their practical deployment is hampered by a critical tradeoff between the high performance of large models and the cost-efficiency of smaller
ones. This research addresses this challenge by designing and evaluating an intelligent
LLM routing system to optimize both cost and performance. Four distinct routing
methods, including standalone classifiers like XGBoost and a hybrid KNN+XGBoost
model, were evaluated on their ability to dynamically select the most suitable LLM for
queries across diverse benchmarks: conversational (MT-Bench), mathematical
reasoning (GSM8K), and multi-domain (MMLU). The results demonstrate that the
proposed routing framework achieves significant cost savings, reducing expenses by
up to 2.76× on MT-Bench with minimal impact on response quality. The hybrid
KNN+XGBoost model proved most robust, particularly on the challenging MMLU
benchmark where other models failed, simpler classifiers struggled in specialized
domains. Ultimately, this project validates LLM routing as a powerful strategy for cost
optimization
Development of a lower-extremities exoskeleton for postural-sway in sit-to-stand movement
Get PDFPostural sway, a critical indicator of balance control, is often heightened in individuals with chronic low back pain (CLBP), age-related decline, and neurodegenerative disorders such as Parkinson’s disease. While passive exosuits offer ergonomic support, they lack the capacity for real-time corrective intervention. This project introduces a lightweight, fabric-based active back support exoskeleton powered by pneumatic actuation, designed to detect and
correct static postural sway. The system integrates an ESP32 microcontroller, IMU-based sway detection, and a threshold-based control algorithm for real time actuation. The fully assembled prototype was evaluated in 15 subjects across varying stance (normal, tandem) and visual (eyes open/closed) conditions using surface electromyography (sEMG) and centre of pressure (CoP) metrics. Under the most challenging balance condition, Tandem Stance Eyes Closed (TSEC), results showed a 51.5% reduction in CoP pathlength, 27.4% and 41.8%
decreases in Vapmean and Vmlmean, and 34.7% and 20.7% reductions in DMLSD and DAPSD, respectively. sEMG analysis indicated a significant drop in trunk muscle activation, with External Obliques (38.3%), Rectus Abdominis (51.2%), and Erector Spinae (L3) Right (41.8%) showing the largest reductions in RMS amplitude during TSEC trials. The sway detection algorithm achieved 70% classification accuracy, supporting low-power, real-time execution on embedded hardware. These findings validate the exosuit’s ability to enhance postural stability and reduce neuromuscular strain during quiet stance. The system demonstrates potential as a practical rehabilitation aid for individuals with CLBP, older adults, or those with early-stage Parkinson’s disease. Future
work may involve adaptive control integration and broader clinical validation.
Keywords: Active Exoskeleton, Postural Sway, Centre of Pressure (CoP), Kinematics, Biomechanics
Subject Area: R856-857 Biomedical engineering. Electronics. Instrumentatio
A machine learning approach to movie recommendation system
Get PDFWith the rapid growth of the digital entertainment industry and the increasing popularity of streaming platforms like Netflix and YouTube, personalized content delivery has become a critical focus. Intelligent recommendation systems are essential for enhancing user engagement, reducing decision fatigue, and promoting content discovery. This project presents a machine learning-based movie recommendation system aimed at providing accurate and personalized movie suggestions while addressing common industry challenges such as biased recommendations, data sparsity, and the cold start problem. Multiple algorithms—including K-Means with KNN, Singular Value Decomposition (SVD), and Matrix Factorization using Keras—were evaluated using Root Mean Square Error (RMSE) to identify the most effective model. A hybrid approach, integrating content-based and collaborative filtering techniques, was adopted to optimize recommendation accuracy and fairness. The final system is implemented as a web application with features such as secure login, dynamic movie interaction, and personalized profile management. This work demonstrates the potential of intelligent systems to improve user satisfaction in digital media platforms
Stock indicator scanner customization tool
Get PDFStock market prediction is a critical yet challenging task due to the inherent volatility and complexity of the stock market. Traditional stock analysis techniques, such as technical and fundamental analysis, often fail to predict price movement due to a lack of trading experience and time. These methods, although valuable, lack the flexibility to adapt to evolving market conditions and personalised trading strategies. Recent advancements in deep learning have shown promise in improving stock trend prediction accuracy. This project seeks to bridge the gap by developing a customisable stock indicator scanner tool that integrates deep learning, specifically Long Short-Term Memory (LSTM) networks, with technical analysis. The proposed system enables users to customise input parameters, such as Simple Moving Average (SMA), Exponential Moving Average (EMA), Moving Average Convergence Divergence (MACD), Relative Strength Index (RSI), Stochastic Oscillator, and 46 other technical indicators. This allows users to personalise the models trained to the user preferences of technical indicators. This customisation enables traders to adjust the model based on their unique trading strategies. Additionally, the system recommends optimal combinations of technical indicators to improve model performance, making the tool adaptive to different market scenarios. The model will be evaluated using accuracy. By incorporating deep learning, this project offers a novel customisation approach to stock trend prediction, enabling investors to make informed decisions with personalised models that better align with their trading goals. This tool has the potential to significantly enhance stock market analysis by providing more accurate and adaptive predictions, contributing to the broader field of financial market forecasting
A gamified e-learning web application with ai-driven study planner utilizing llama
Get PDFE-learning has transformed the way of traditional learning by making learning accessible for students anywhere and anytime with technology devices. Thus, the project focuses on developing a gamified e-learning web application powered by artificial intelligence (AI) using Large Language Models (LLMs) integrated with LangChain for secondary school students to improve the learning experience. The web application is built using the Model-View-Controller (MVC) architecture, with Laravel for the backend and React for the front end. The three user modules, students, teachers, and administrators, are included to provide role-specific functionalities. Students can enroll in courses, access learning materials, engage in gamified challenges, track their learning progress, and participate in discussion forums. They also benefit from personalized AI-generated study plan timetables that analyze their strengths and weaknesses across subjects. Experimental results show that LLaMA 3.1 achieves higher accuracy and stability than Qwen 2.5, making it a stronger candidate for AI-driven study planning. Teachers are able to manage course content, monitor student performance, provide feedback, and interact with students, while administrators oversee user management, platform usage, and gamification elements. By combining gamification with AI-driven personalization, the web application aims to motivate students for more engagement to improve learning retention.
Keywords: Artificial Intelligence, E-Learning, Gamification, Large Language Models (LLMs), LLaMA, Personalized Learning
Subject Area: AZ191-193 Evaluation, LB1603-1696.6 Secondary education. High schools, T61-173 Technical education. Technical school
晏殊《珠玉词》“伤春”主题探究 - An exploration of the "Melancholy for Spring" theme in Yan Shu's Jade Lyrics
No full text晏殊《珠玉词》词风典雅清丽，犹以表达春愁、伤春情怀见长，展现出晏殊其 人对生命与人生无常的哲思。晏殊作为北宋词坛奠基性人物，其词作多描绘春景、寄 托情思，展现出浓厚的“伤春”意蕴，形成其特有的婉约风格与抒情方式。本文以 “伤春”为核心主题，聚焦其词中对季节流转与人生无常的慨叹，透过具象化的意象 与深层情感建构出独特的审美意境。本研究将探析其中“伤春”主题的文化渊源、意 象表现、情感层次与艺术手法。“伤春”作为中国古典文学中的重要抒情传统，在晏 殊的词作中表现出富贵闲愁、时光易逝、离情别绪等多重情感维度，兼具审美性与哲 理性，反映宋代文人感慨人生无常与追求从容洒脱的心理状态。研究首先界定“伤春” 的主题与传统文化根源，指出此类情感在继承了唐人“惜春”传统的同时，又在词中 发展为更具主观情感色彩的抒怀方式。随后，笔者将《珠玉词》中二十八首含有“伤 春”意蕴的作品依照时间意象分类，继而将晏殊的词作分为早春、盛春、暮春三种 “伤春”类型，进一步从自然意象如花、柳、燕、风雨等角度入手，解析其如何融情 于景、借物言志，构建多重情境。通过分析可见，晏殊善用对仗、衬托、拟人等手法， 使词作充满抒情张力。同时，文章运用细读法深入剖析其婉约含蓄的语言风格与情景 交融的修辞方式，揭示晏殊是如何以优雅词章表现生命的短暂与人世的无常。最后结 合叶嘉莹《灵谿词说》中相关评价与现代学者对晏词的评价，指出晏殊“伤春”不仅 体现宋代士大夫的文化心态，也彰显了其在词中的独特地位与艺术高度。晏殊的“伤 ix 春”至此不仅展现其个人情思，也隐含他对世事无常、韶华易逝的体认，本文旨在分 析晏殊“伤春”主题的多样性与多维意涵，展现其词中“情”与“理”的独特交织。 【关键词】1. 晏殊；2. 《珠玉词》；3. 伤春主题；4. 婉约词风；5.宋词美
金庸《射雕英雄传》中黄药师之“邪”研究 : A study on <The Legend of the Condor Heroes> in the evil of Huang Yaoshi
No full text金庸武侠小说中除了颇为引人注目的主人公们，其中还隐藏许多独特个性的配角们， 以他们的独有色彩来点缀武侠小说中的江湖世界。金庸除了在塑造角色的手法上有所讲究， 对于每一个角色的独特性，更是耗费不少心血与智慧，才能使每一位角色在武侠小说中栩 栩如生。金庸笔下武功高强的角色众多，唯有一人不免被人察觉他的独特，就是——黄药 师。别称“东邪”。金庸在《射雕英雄传》中描写不少关于黄药师“邪”的形象，更是添 加他的“正”来形成他的矛盾型人物特征。 本文主要针对《射雕英雄传》中黄药师的人物形象作为研究角度，进行分析他的“邪” 与“正”。第一章为绪论，说明本文的研究目的、范围、方法、前人研究述评以及论文框 架。第二章为“黄药师的人物性格与形象”，探讨其性格对情节发展的作用性以及独特之 处。第三章为“黄药师武学之天人合一”，此章结合道家思想分析其武学的描写与命名。 第四章为“黄药师之“邪中带正””，批判黄药师的“邪”以及论述他的“正”，并借以 《神雕侠侣》作为补充。本文的最后一章为结语，总结以上研究内容，并给予论点与研究 成果。 【关键词】黄药师、金庸、射雕英雄传、人物形象、武
Correlation between frequency of ChatGPT usage and academic writing motivation of ESL learners
Get PDFIn the evolving landscape of academic communication, digital tools have emerged as pivotal companions, particularly for ESL learners navigating the complex terrain of academic writing. This study explores the perceived effectiveness and specific contributions of digital writing tools, namely ChatGPT and Grammarly, among 106 undergraduate ESL learners at Universiti Tunku Abdul Rahman (UTAR). Employing a mixed-methods design, quantitative data were collected via structured questionnaires, while qualitative insights were drawn from in-depth semi-structured interviews. Findings reveal that students overwhelmingly recognised the tools’ utility in enhancing grammar accuracy, vocabulary sophistication, coherence, and organisational structure. Real-time feedback and instant assistance were identified as key features that bolstered both writing efficiency and learner confidence. However, limitations such as over-reliance, lack of contextual nuance, and concerns over citation accuracy surfaced, underscoring the need for critical engagement and guided usage. The study not only affirms the transformative potential of digital tools in scaffolding academic writing development but also highlights the necessity for localised pedagogical support to ensure that these digital tools empower rather than replace the learner’s voice. In illuminating the intersection of digital tools and ESL academic writing, this research contributes timely insight into the digital shift reshaping language education in Malaysian private higher institutions
AI for pest detection
Get PDFAgricultural pest management is critical for ensuring crop health and yield, yet traditional detection methods are often labor-intensive and imprecise. This Final Year Project proposes an Artificial Intelligence (AI) based pest detection system integrated into a mobile application, empowering users to monitor plant health efficiently. The system utilizes a convolutional neural network (CNN) specificallyYOLOv5, trained on a dataset encompassing three common pest categories tentatively Whiteflies, and caterpillar, alongside healthy plant samples. Through the mobile app, users capture images of plants using their smartphone camera. The AI model, running on-device through TensorFlow Lite, analyzes the image to classify the plant as healthy or unhealthy. If unhealthy, it identifies the specific pest type and provides a tailored solution. Results are displayed within the app, and the detection data including plant status, pest type, solution, and confidence score are encoded into a QR code. This QR code enables seamless data sharing, such as with agricultural experts or record-keeping systems. Preliminary testing on a diverse test set achieved an accuracy above 85%, validating the system’s effectiveness. This mobile solution offers a portable, user-friendly tool for pest management, enhancing precision agriculture through AI and innovative data transfer