School of Engineering, Computer and Mathematical Sciences - Te Kura Mātai Pūhanga, Rorohiko, Pāngarau

Permanent link for this collectionhttps://hdl.handle.net/10292/553

AUT is home to a number of renowned research institutes in engineering, and computer and mathematical sciences. The School of Engineering, Computer and Mathematical Sciences strong industry partnerships and the unique combination of engineering, computer and mathematical sciences within one school stimulates interdisciplinary research beyond traditional boundaries. Current research interests include:

Artificial Intelligence; Astronomy and Space Research;

Biomedical Technologies;

Computer Engineering; Computer Vision; Construction Management;

Data Science;

Health Informatics and eHealth;

Industrial Optimisation, Modelling & Control;

Information Security;

Mathematical Sciences Research; Materials & Manufacturing Technologies;

Networking, Instrumentation and Telecommunications;

Parallel and Distributed Systems; Power and Energy Engineering;

Software Engineering; Signal Processing; STEM Education;

Wireless Engineering;

Browse

Now showing 1 - 20 of 1766

Development of a Cost-Effective UUV Localisation System Integrable with Aquaculture Infrastructure
(MDPI, 2026-01-07) Huang, Loulin; Tun, Thein Than; Preece, Mark Anthony
In many aquaculture farms, Unmanned Underwater Vehicles (UUVs) are being deployed to perform dangerous and time-consuming repetitive tasks (e.g., fish net-pen visual inspection) on behalf of or in collaboration with farm operators. Mostly, they are remotely operated, and one of the main barriers to deploying them autonomously is the UUV localisation. Specifically, the cost of the localisation sensor suite, sensor reliability in constrained operational workspace and return on investment (ROI) for the huge initial investment on the UUV and its localisation hinder the R&D work and adoption of the autonomous UUV deployment on an industrial scale. The proposed system, which leverages the AprilTag (a fiducial marker used as a frame of reference) detection, provides cost-effective UUV localisation for the initial trials of autonomous UUV deployment, requiring only minor modifications to the aquaculture infrastructure. With such a cost-effective approach, UUV R&D engineers can demonstrate and validate the advantages and challenges of autonomous UUV deployment to farm operators, policymakers, and governing authorities to make informed decision-making for the future large-scale adoption of autonomous UUVs in aquaculture. Initial validation of the proposed cost-effective localisation system indicates that centimetre-level accuracy can be achieved with a single monocular camera and only 10 AprilTags, without requiring physical measurements, in a 115.46 m3 laboratory workspace under various lighting conditions.
Ensuring Reliable District Heating Systems: Identifying Critical Components under Independent and Cascading Failure Scenarios
(Elsevier BV, 2026-01-05) Mao, Ding; Xu, Sai; Wang, Jay; Shen, Linhua; He, Wei
Urban district heating systems are vital infrastructures of sustainable cities, providing efficient and centralized thermal energy to residential and industrial users. However, these systems consist of numerous interdependent components that are prone to faults, which can disrupt heat supply and compromise service reliability. Identifying critical components to maintain system stability is crucial for enhancing the resilience and sustainability of urban energy infrastructure. Critical components are generally determined by evaluating the consequences of failures, which involves simulating all possible fault scenarios, a process that is computationally expensive and time-consuming. To address this challenge, we propose a comprehensive component importance identification framework. This framework incorporates two methods: the Importance Calculation Method (ICM), which operates under normal system conditions, and the Failure-Simulation-Based Method (FSM), which simulates failure consequences. These methods evaluate component criticality under both independent and cascading failure scenarios, incorporating topological and functional perspectives. To validate the proposed framework, gridded heating system models of varying scales, comprising 4-, 9-, 16-, and 25-node configurations, were developed. Applying the framework to these models revealed a strong correlation between ICM and FSM results: the topological importance index in ICM showed a high correlation with FSM’s functional consequence indices (ρ > 0.75), while the functional importance indices achieved even higher correlations (ρ = 0.94–0.97). Finally, the framework was applied to a real-world district heating system in China, where it successfully identified critical pipes and demonstrated the effectiveness and practical value of the proposed ICM through comparison with traditional fault-simulation-based methods.
Optimal Access Structure Partition Methods for Image Secret Sharing
(Institute of Electrical and Electronics Engineers (IEEE), 2025-12-19) Wu, Xiaotian; Tang, Li; Xia, Zhihua; Yang, Ching-Nung; Yan, Weiqi
Visual cryptography scheme (VCS) and polynomial-based secret image sharing (PSIS) are two primary types of secret sharing for protecting images. VCS and PSIS have their respective pros and cons. For VCS, the benefits of perfect security and easy decoding are provided. But it suffers from the limitations of lossy secret recovery and binary image-oriented. PSIS can deal with grayscale/color images and offers lossless secret reconstruction. Whereas, the secret decoding is computationally intensive (i.e., O(klog²k) for (k,n) threshold) and the residual-image problem in PSIS compromises the security. In this paper, we are motivated to investigate a sharing technique that can preserve the advantages of both VCS and PSIS. Differing from existing VCS and PSIS, the proposed sharing method is accomplished based on the access structure partition (ASP) result. Essentially, an ASP guided image secret sharing approach is developed and three optimal ASP algorithms are designed. When compared with existing partition method, significant improvement is offered by our partition techniques especially for the (k,n) threshold with a larger n . Take the (2,15) , (2,18) , and (4,12) thresholds for example, the numbers of involved sub-access structures by our method are 4, 5, and 19, while the quantities by existing approach are 8, 10, and 45. The percentages of improvement are 100%, 100%, and 137%. Further, based on the partition result from ASP algorithms, we can employ (k,k) probabilistic VCS (PVCS) to constitute a (k,n) sharing method for encoding gray-level/color images. Experiments are demonstrated to confirm the effectiveness of the sharing method and ASP algorithms. Meanwhile, comparisons are included to show that the merits of perfect security, low decoding complexity (i.e., O(d) ), lossless secret recovery (i.e., PSNR =∞ , SSIM= 1), and grayscale/color image-oriented are provided by our sharing method.
Interactive Visualisation of Complex Street Network Graphs from OSM in New Zealand
(MDPI AG, 2025-12-07) Ng, Jun Yi; Ma, Jing; Singh, Anuradha; Lai, Edmund M-K; Hayman, Steven
Street network graphs model interconnected land transport infrastructure, including roads and intersections, enabling traffic analysis, route planning, and network optimization. Directed network graphs (digraphs) add directionality to these connections, reflecting one-way streets and complex traffic flows. While OpenStreetMap (OSM) offers extensive data, visualizing large-scale directed networks with complex junctions remains computationally challenging for browser-based tools. This paper presents an interactive visualization tool integrating OSM data with the New Zealand Transport Agency’s National Network Performance (NNP) analysis toolbox using PyDeck and WebGL. We introduce a directional offset algorithm to resolve edge overlaps and a geometry-aware node placement method for complex intersections. Experimental results demonstrate that our PyDeck implementation significantly outperforms existing solutions like Bokeh and OSMnx. On standard datasets, the system achieves up to 238× faster processing speeds and a 93% reduction in output file size compared to Bokeh. Furthermore, it successfully renders metropolitan-scale networks (∼1.3 million elements) where traditional visualisation tools fail to execute. This visualisation approach serves as a critical debugging instrument for NNP, allowing transport modellers to efficiently identify connectivity errors and validate the structural integrity of large-scale transport models.
A Stacked Substrate-Integrated Waveguide-Based Pyramidal Horn Antenna for Terahertz Communications
(MDPI AG, 2025-12-04) Paudel, Biswash; Li, Xue Jun; Seet, Boon-Chong
The terahertz (THz) band offers ultra-wide bandwidth for next-generation high-speed wireless communication systems. However, achieving compact, high-gain, and beam-symmetric THz antennas remains challenging due to fabrication and propagation constraints. This paper presents a simulation-based design and optimization of a stacked substrate-integrated waveguide (SIW) pyramidal horn antenna achieving equal half-power beamwidths (HPBWs) in both E- and H-planes. The design employs vertically stacked SIW layers coupled through optimized slot apertures to ensure dominant TE10 mode propagation with minimal reflection. Using full-wave electromagnetic simulations, the effects of layer number, dielectric loading, amplitude tapering, and phase distribution are systematically analyzed. The optimized five-layer configuration exhibits 10 dBi gain, 41° HPBW, and sidelobe levels around −3.2 dB at 210 GHz. This framework aims to develop high-performance, beam-symmetric THz SIW antennas compatible with standard LTCC/PCB technologies.
Analysis of the Effect of Using a Variable Speed Drive on the Power Consumption of the ID Fan Drive Motor
(EDP Sciences, 2025-11-20) Kastawan, I Made Wiwit; Amanda, Rizki; Mulyadi, Ahmad Deni; Murniyati, Dyah Ayu Yuli; Zamora, Ramon
Cirebon Coal-Fired Power Plant Unit 1 uses an Induced Draft Fan (ID Fan) to regulate flue gas flow and maintain negative pressure inside the boiler. The current ID Fan control system still applies a blade pitch position mechanism with constant motor speed, resulting in high power consumption ranging from 2,547.9 kW to 4,311.2 kW. This condition is inefficient because the motor runs at constant speed regardless of changing load demands. This study aims to design a Variable Speed Drive (VSD) to control the ID Fan motor speed and analyze its impact on power consumption. The research was conducted using simulation with MATLAB/Simulink R2021a software. The VSD design consists of three main components: a three-phase rectifier, a buck converter-based DC link, and a three-phase PWM inverter, adjusted to meet the ID Fan operational pressure limit of 8.0 to 10.0 kPa. The simulation results show that with the implementation of VSD, the system pressure can be maintained within the safe range of 8.5 to 9.7 kPa, and the motor speed can be adjusted according to airflow demand. Power consumption decreased from 2,645.7 kW to 1,273.1 kW after implementing VSD, resulting in an energy saving of 62.1%. The application of VSD is proven to be effective in improving energy efficiency in the ID Fan system at the Cirebon Coal-Fired Power Plant Unit.
Unsupervised Thematic Context Discovery for Explainable AI in Fact Verification: Advancing the CARAG Framework
(SciTePress - Science and Technology Publications, 2025-12) Vallayil, Manju; Nand, P; Yan, Wei Qi; Allende-Cid, H
This paper introduces CARAG-u, an unsupervised extension of the Context-Aware Retrieval Augmented Generation (CARAG) framework, designed to advance explainability in Automated Fact Verification (AFV) architectures. Unlike its predecessor, CARAG-u eliminates reliance on predefined thematic annotations and claim-evidence pair labels, by dynamically deriving thematic clusters and evidence pools from unstructured datasets. This innovation enables CARAG-u to balance local and global perspectives in evidence retrieval and explanation generation. We benchmark CARAG-u against Retrieval Augmented Generation (RAG) and compare it with CARAG, highlighting its unsupervised adaptability while maintaining a competitive performance. Evaluations on the FactVer dataset demonstrate CARAG-u’s ability to generate thematically coherent and context-sensitive post-hoc explanations, advancing Explainable AI in AFV. The implementation of CARAG-u, including all dependencies, is publicly available to ensure reproducibility and support further research.
Bayesian Estimation of R-Vine Copula with Gaussian-Mixture GARCH Margins: An MCMC and Machine Learning Comparison
(MDPI AG, 2025-12-04) Khanthaporn, Rewat; Wichitaksorn, Nuttanan
This study proposes Bayesian estimation of multivariate regular vine (R-vine) copula models with generalized autoregressive conditional heteroskedasticity (GARCH) margins modeled by Gaussian-mixture distributions. The Bayesian estimation approach includes Markov chain Monte Carlo and variational Bayes with data augmentation. Although R-vines typically involve computationally intensive procedures limiting their practical use, we address this challenge through parallel computing techniques. To demonstrate our approach, we employ thirteen bivariate copula families within an R-vine pair-copula construction, applied to a large number of marginal distributions. The margins are modeled as exponential-type GARCH processes with intertemporal capital asset pricing specifications, using a mixture of Gaussian and generalized Pareto distributions. Results from an empirical study involving 100 financial returns confirm the effectiveness of our approach.
A Secure and Sustainable Transition from Legacy Smart Cards to Mobile Credentials in University Access Control Systems
(MDPI AG, 2025-12-04) Mustafa, Rashid; Khan, Toseef Ahmed; Sarkar, Nurul I
A secure and sustainable building access control system plays a vital role in protecting organisational assets worldwide. Physical access management at Auckland University of Technology (AUT) is still primarily done through traditional card-based authentication. The system is susceptible to replay and cloning attacks because the conventional Mifare Classic credentials employ outdated Crypto1 encryption. Such weaknesses provide significant threats in laboratories, engineering testing facilities, and research and technological areas that require strict security procedures. To overcome the above issues, we propose a secure and sustainable university building access control system using mobile app credentials. This research grounded a thorough risk analysis of the university’s current infrastructure, mapping potential operational continuity threats. We analyse card issuance records by identifying high-risk areas such as restricted laboratories and evaluating the resilience of the current Gallagher–Salto system against cloning and replay attacks. We quantify the distribution and usage of cards that are vulnerable. To evaluate the risks to operational continuity, the system architecture is examined. Additionally, a trial implementation of the Gallagher Mobile Connect platform was conducted, utilising cloud registration, multi-factor authentication (PIN or biometrics), and books. Pilot implementation shows that mobile-based credentials improve user experience, align with AUT’s environmental sustainability roadmap, and increase resilience against known attacks. Results have shown that our proposed mobile credentials can improve the system performance up to 80%.
AI-Driven Energy-Efficient Routing in IoT-Based Wireless Sensor Networks: A Comprehensive Review
(MDPI AG, 2025-12-05) Thakur, Sumendra; Sarkar, Nurul I; Yongchareon, Sira
Efficient routing remains the linchpin for achieving sustainable performance in Wireless Sensor Networks (WSNs) within the Internet of Things (IoT). However, traditional routing mechanisms increasingly struggle to cope with the growing complexity of network architectures, frequent changes in topology, and the dynamic behavior of mobile nodes. These issues contribute to data congestion, uneven energy consumption, and potential communication breakdowns, underscoring the urgency for optimized routing strategies. In this paper, we present a comprehensive review of over 100 studies of spanning conventional and AI-enhanced energy-efficient routing techniques. It covers diverse approaches, including metaheuristics, machine learning, reinforcement learning, and AI-based cross-layer methods aimed at improving the performance of WSN-IoT systems. The key limitations of existing solutions are discussed along with performance metrics such as scalability, energy efficiency, throughput, and packet delivery. We also highlight various research challenges and provide research directions for future exploration. By synthesizing current trends and gaps, we provide researchers and practitioners with a structured foundation for advancing intelligent, energy-conscious routing in next-generation IoT-enabled WSNs.
Perspective on the Role of AI in Shaping Human Cognitive Development
(MDPI AG, 2025-11-20) Abbosh, Amin; Al-Anbuky, Adnan; Xue, Fei; Mahmoud, Sundus S
The fourth industrial revolution, driven by Artificial Intelligence (AI) and Generative AI (GenAI), is rapidly transforming human life, with profound effects on education, employment, operational efficiency, social behavior, and lifestyle. While AI tools potentially offer unprecedented support in learning and problem-solving, their integration into education raises critical questions about cognitive development and long-term intellectual capacity. Drawing parallels to previous industrial revolutions that reshaped human biological systems, this paper explores how GenAI introduces a new level of abstraction that may relieve humans from routine cognitive tasks, potentially enhancing performance but also risking a cognitively sedentary condition. We position levels of abstraction as the central theoretical lens to explain when GenAI reallocates cognitive effort toward higher-order reasoning and when it induces passive reliance. We present a conceptual model of AI-augmented versus passive trajectories in cognitive development and demonstrate its utility through a simulation-platform case study, which exposes concrete failure modes and the critical role of expert interventions. Rather than a hypothesis-testing empirical study, this paper offers a conceptual synthesis and concludes with mitigation strategies organized by abstraction layer, along with platform-centered implications for pedagogy, curriculum design, and assessment.
Understanding Security Vulnerabilities in Private 5G Networks: Insights from a Literature Review
(MDPI AG, 2025-10-23) Fue, Jacinta; Gutierrez, Jairo A; Donoso, Yezid
Private fifth generation (5G) networks have emerged as a cornerstone for ultra-reliable, low-latency connectivity across mission-critical domains such as industrial automation, healthcare, and smart cities. Compared to conventional technologies like 4G or Wi-Fi, they provide clear advantages, including enhanced service continuity, higher reliability, and customizable security controls. However, these benefits come with new security challenges, particularly regarding the confidentiality, integrity, and availability of data and services. This article presents a review of security vulnerabilities in private 5G networks. The review pursues four objectives: (i) to identify and categorize key vulnerabilities, (ii) to analyze threats that undermine core security principles, (iii) to evaluate mitigation strategies proposed in the literature, and (iv) to outline gaps that demand further investigation. The findings offer a structured perspective on the evolving threat landscape of private 5G networks, highlighting both well-documented risks and emerging concerns. By mapping vulnerabilities to mitigation approaches and identifying areas where current solutions fall short, this study provides critical insights for researchers, practitioners, and policymakers. Ultimately, the review underscores the urgent need for robust and adaptive security frameworks to ensure the resilience of private 5G deployments in increasingly complex and high-stakes environments.
A Bibliographic Study of Integrating IoT and Geospatial Modelling for Sustainable Smart Agriculture in Developed Countries: Focus on Australia
(Elsevier, 2025-12-02) Mamun, Quazi; Zaman, Asaduz; Ip, Ryan HL; Haque, KM Shamsul
Integrating the Internet of Things (IoT) and geospatial modelling technologies is pivotal for advancing sustainable smart agriculture, particularly in resource-constrained environments like Australia. This systematic literature review examines the adoption and impact of these technologies in agriculture across Australia and select developed countries. Through an extensive analysis of 172 peer-reviewed articles published between 2013 and 2023, this study identifies key technological advancements such as unmanned aerial vehicles (UAVs), consumer-grade cameras (RGB cameras), and satellite platforms (Sentinel-2, LANDSAT-8) that have significantly influenced agricultural practices. The findings reveal Australia’s progress in adopting these technologies but also highlight gaps compared to countries like Germany and the USA, especially in using UAVs, Synthetic Aperture Radar (SAR) and RGB cameras. The study underscores Australia’s need to enhance its technological capabilities, particularly resource management, to foster more efficient and sustainable agricultural practices. This review provides valuable insights for policymakers, researchers, and technology providers, aiming to drive innovation and improve agricultural outcomes in the face of growing environmental challenges.
Ethical and Societal Impacts of Generative AI in Higher Computing Education: An ACM Task Force Working Group to Develop a Landscape Analysis – Perspectives from the Global Souths and Guidelines for CS1/CS2/CS3
(ACM, 2025-06-17) Szabo, C; Sheard, J; Dake, DK; Falkner, NJG; Enock, M; Ogunyemi, O; Mbodila, M; Clear, T; Ola, O; Taukobong, T; Wadhwa, B
Generative AI has a wide range of impacts on how we access and use information, particularly as educational settings and perspectives differ greatly across different locations. These impacts extend to society and include impacts on intellectual and creative works and the potential infringement of authorship. Differences in institutional GenAI policies (and in funding) may create unequal access to AI tools, the potential disparity in student knowledge of AI tools, responsible uses of AI tools, ethical questions about AI tools, and uneven student knowledge of the benefits and limitations of AI tools. Generative AI introduces questions concerning academic integrity, bias, and data provenance. The training data’s source, reliability, veracity, and trustworthiness may be in doubt, creating broader societal concerns about the output of the Generative AI models. This working group will conduct a landscape analysis on Global South ethical questions related to the use of Generative AI tools in higher education contexts, identifying promising principles, challenges, and ways to navigate the implementation of Generative AI in ethical and principled ways.
An Experimental EACI-Based Localization Framework Using LQI and CNN for Consumer IoT
(Institute of Electrical and Electronics Engineers (IEEE), 2025-11-11) Ahmad, T; Hadi, MU; Li, XJ; Anwar, A; Ibrahim, MM; Khan, S
Precise indoor localization remains a challenge in wireless sensor networks (WSNs) due to multipath fading, interference, and signal fluctuations in different environments. Traditional methods depend on Received Signal Strength (RSS) also often struggle with accuracy in indoor scenario. This study presents an experimental localization framework that utilizes Link Quality Indicator (LQI) values and Convolutional Neural Networks (CNNs) within an Edge Computing-Assisted Consumer IoT (EACI) model. The proposed approach segments the network using a pyramid-loop algorithm and employs LQI-based measurements for more stable and accurate distance estimation. A CNN classifier is trained on normalized LQI data, including statistical features such as kurtosis, to predict node locations. The system is authenticated by a real-world testbed using Zigbee XB24C nodes. The experimental results show an overall localization error of 0.12m at zone 1 with a standard deviation of 0.89m. This reflects an improved localization accuracy and reduced error compared to RSS-based and existing CNN-based methods. The proposed technique effectiveness is observed for indoor localization in consumer IoT environments.
An Alternative Cardinal Spline for Cubic B-Spline Interpolation
(IEEE, 2025-11-20) Chen, J; Li, Q; Chen, J; Yan, Wei Qi
This paper aims at introducing a special cubic cardinal spline from the GMK-splines, which is able to obtain almost the same interpolation results of the cubic B-spline, but without solving the system of equations for getting control vertices, and exploring its applications in geometric modeling and image processing. The spline involved can be obtained by using a linear combination of several shifted B-splines of the same degree. The spatial and frequency domain comparison demonstrates its superior local support and frequency domain performance. In addition, a number of examples are involved in geometric modeling and digital image processing, compared with a few conventional methods. As shown in this paper, the cubic cardinal spline is an alternative choice instead of the cubic B-spline in the interpolation procedure in order to avoid solving the system of equations over and over again.
Diffusion Model for A Virtual Try-On System
(IEEE, 2025-12-21) Zhang, Yuchao; T. P. Tran, Kien; Nguyen, Minh; Yan, Wei Qi
We present a modular virtual try-on (VTON) system that integrates natural language control, efficient diffusion-based image synthesis, and lightweight garment classification. User intent is parsed by a large language model (LLM) into structured visual prompts. A LoRA-tuned diffusion model generates tryon images conditioned on pose and segmentation maps, while a compact classifier, LightClothNet, handles five-category clothing recognition and pre-filtering. The pipeline is built using ComfyUI nodes and orchestrated via Dify. Compared to the existing methods, the proposed system offers improved realism, garment-pose alignment, and controllability. Our evaluations on the DressCode and VITON-HD datasets show that LoRA fine-tuning enhances fidelity under limited data, while LightClothNet achieves up to 91.76% precision and 0.91 F1-score with low latency. This result demonstrates how multimodal control, lightweight classification, and diffusion generation are unified for fast, flexible, and userdriven VTON applications.
Computational Analysis and Statistics of Table Tennis Games
(IEEE, 2025-12-21) Guangliang, Yang; Nguyen, Minh; T. P. Tran, Kien; Yan, W
This paper presents ChatPPG prototype, which is an innovative system that combines large language models (LLMs) fine-tuned with Low-Rank Adaptation (LoRA) and computer vision for real-time data analysis and coaching for table tennis games. By integrating multi-camera 3D reconstruction, visual object detection and object tracking, ChatPPG processes match data such as player speed, ball trajectories, and service legality, transforming raw metrics into actionable insights. The fine-tuned model achieved a Q/A accuracy 92.3 %, surpassing the baseline model 83.7 %, with sub-second response times enabled by 8-bit quantization. Practical applications demonstrated its ability to deliver personalized training plans and tactical recommendations tailored to individual player profiles. User feedback from professional coaches and athletes rated tactical suggestions at 9.3/10 and training recommendations at 8.9/10. Integrating structured CV outputs with LLM capabilities enhanced transparency and interpretability, allowing users to trace recommendations to data-driven decisions. Despite dataset limitations and the need for advanced query handling, ChatPPG bridges the gap between data analysis and decision-making, setting a new standard for integrating LLMs and CV technologies in fast-paced sports analytics.
Contrasting Big Data Techniques in Exploring New Zealand Road Crash Data
(Auckland University of Technology, 2025-11-17) Thorpe, Stephen; Hu, Baosen (Edison)
Motor vehicle crashes result in high social and economic costs globally and in New Zealand. Therefore, accurate analysis of crash events is critical for evidence-based prevention and policy. This study explored the application of Big Data techniques, specifically Hadoop and MapReduce, to improve the analysis of the impact of weather and speed on motor vehicle crashes in New Zealand. Contemporary Big Data approaches were applied to address the limitations inherent in traditional methods of crash analysis. We used Hadoop’s distributed storage and MapReduce’s processing capabilities on the New Zealand Transport Agency’s Crash Analysis System (CAS) dataset to identify and visualize environmental and spatial trends to a higher degree of understanding. The project involved Elasticsearch and Kibana to make sense of unstructured data in geographic views, while Hue, Hive, and Power BI represented structured data with charts and dashboards. Results show that non-injury crashes, followed by minor crashes, are the most frequent, with over half happening at speed limits between 40–60 km/h. Geographically, Auckland represents crashes five times greater than in the other locations. Strong and extreme weather conditions appear to be a factor in the majority of reported fatal road accidents.
Improving Story Points Estimation Using Ensemble Machine Learning
(Springer Science and Business Media LLC, 2025-11-13) Ahmad, Z; Kuo, MMY
Agile software development (ASD) emphasizes iterative development, continuous feedback, and team collaboration, addressing the limitations of traditional methodologies. This research explores the application of machine learning (ML) to improve story point estimation in ASD, a critical practice for planning and prioritization. Traditional methods like Planning Poker often suffer from human biases and inconsistencies, leading to unreliable estimates. This study introduces an innovative ML-based ensemble stacking technique, combining RoBERTa, a transformer model for natural language processing, with BiLSTM, a neural network adept at handling sequential data. The research involves reviewing existing ML methodologies, developing the proposed model, and evaluating its effectiveness using 21,064 data points from 14 open-source projects. The model’s performance was assessed through Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). Results show that the proposed ensemble model achieved lower MAE and MAPE, with performance improvements ranging from 4% to 32% over state-of-the-art models. While promising, the study suggests there is still room for further refinement, indicating the potential for ongoing advancements. This research contributes to the integration of ML in software engineering, offering a path toward more accurate and efficient project management.

Browse

Recent Submissions