AUT Research Repository (Auckland Univ. of Technology)
Not a member yet
14707 research outputs found
Sort by
Integrating Explainable Artificial Intelligence (AI) to Evaluate Fermented-Roasted Avocado Seed Powder as a Natural Antioxidant in Beef Patties
No full textThis study evaluates fermented–roasted avocado seed powder (ASP) as a sustainable antioxidant for beef patties and integrates explainable artificial intelligence (AI) to provide mechanistic insights into lipid oxidation control. ASP was produced via controlled fermentation with Lactobacillus plantarum followed by roasting and incorporated into beef patties (0.95 g/kg). Its performance was compared with butylated hydroxytoluene (BHT) and untreated controls during 10-day refrigerated storage at 4 °C. ASP exhibited the highest initial reducing antioxidant capacity (0.283 mg ascorbic acid equivalents (AAE)/g) measured by cupric reducing antioxidant capacity (CUPRAC) and maintained significantly greater activity than BHT and control throughout storage (p < 0.05). Thiobarbituric acid reactive substances (TBARS) analysis showed ASP reduced lipid oxidation by 21% relative to control, approaching BHT efficacy. ASP also improved colour stability and preserved key fatty acids, notably oleic acid. Volatile profiling revealed distinct antioxidant behaviour, with ASP generating Maillard-derived compounds rather than relying on synthetic additives. Shapley Additive Explanations (SHAP) analysis, applied to an extreme gradient boosting model (XGBoost), identified γ-linolenic acid, oleic acid, pentanal, and 2-heptanone as major predictors of oxidative stability, highlighting ASP's broad-spectrum protective effect. These findings demonstrate that ASP not only valorises avocado seed waste but also offers a viable alternative to synthetic antioxidants, supporting sustainability goals. The integration of explainable AI with multivariate analysis provides new understanding of lipid oxidation pathways and antioxidant performance in meat systems
Frequency-aware Spatio-temporal Topology Learning for Skeleton-based Human Activity Recognition
No full textSkeleton-based human activity recognition (HAR) has made significant progress through graph convolutional networks (GCNs) and Transformer architectures for spatiotemporal modeling. However, existing methods either employ predefined static graph topologies that cannot adapt to heterogeneous skeleton data or learn dynamic topologies based solely on local spatiotemporal features, thereby overlooking the global temporal frequency features of joint movements that are important for discovering semantically meaningful spatial relationships. We propose Frequency-Aware Topology Learning Graph Convolutional Network (FATL-GCN), a novel architecture that integrates frequency-aware temporal context to guide adaptive learning of spatial topology. Our approach leverages Time-to-Vector linear frequency encoding to capture both periodic and non-periodic motion patterns, employs frequency-guided topology learning to generate action-specific graphs through temporal-context-driven attention, and incorporates hierarchical multi-scale fusion for robust feature extraction across scales. Extensive experiments achieved top-1 accuracies of 93.8% (cross-subject) and 97.5% (cross-view) on NTU-60, 91.9% (cross-subject) and 93.1% (cross-setup) on NTU-120, and 51.7% on Kinetics-Skeleton. Ablation studies confirm the critical role of our components, with removing the dynamic graph topology causing a 3.5% accuracy drop and removing frequency-aware encoding causing a 2.1% drop
Circular Supply Chain Design for Biohydrogen Recovery From Perishable Agri-Food Waste
No full textThe increasing interdependencies between water, energy, and food systems highlight the urgency of integrated solutions for managing environmental and resource challenges. This study proposes a sustainable logistics framework for converting agri-food waste into biohydrogen, drawing on the Water-Energy-Food (WEF) Nexus to guide strategic planning. Focusing on Razavi Khorasan, Iran, a drought-prone region with substantial upstream food losses and declining groundwater reserves, the research explores how circular supply chain can support both waste reduction and clean energy generation. The proposed system is structured around a closed-loop supply chain that incorporates both forward delivery and reverse logistics to collect perishable food waste and redirect it for biohydrogen production. This approach prioritizes the recovery of high-water-footprint items such as fruits, vegetables, and cereals, thereby mitigating the loss of embedded resources. A scenario-based assessment of vehicle types and environmental policies highlights the operational and environmental trade-offs of different logistics strategies. The findings suggest that low-capital interventions, such as smart routing and shared logistics, can deliver significant environmental benefits without the infrastructure barriers of full fleet electrification. Ultimately, the framework supports resilient, low-carbon pathways for agri-food systems in water-stressed regions, contributing to circular economy goals and Sustainable Development Goals (SDGs) related to climate action, food security, and clean energy access
Customising Chatbots for Writing Development: Anticipating Semiotic Mediation With the Theoretical Architecture of Systemic Functional Linguistics
No full textThis conceptual paper provides a social semiotic perspective on GenAI technologies in English for Academic Purposes contexts. It focuses on the process of customising AI chatbots to steer how an LLM responds. Through discussing two customised chatbots for Master's of Nursing Science students who are writing research proposals, the paper argues that the theoretical framework of Systemic Functional Linguistics is ideal for chatbot design. Examples use Cogniti software to show how EAP teachers can custom design a chatbot with minimal coding. These examples illustrate how SFL informs decisions about the scope of customised chatbots and the metalanguage within system messages. The discussion of system messages focuses on the challenge of creating consistency with how customised chatbots identify and describe the function of language features when generating feedback messages. The paper argues that this metalanguage should correspond to the metalanguage which students experience in face-to-face teaching and learning as well as online materials. Such continuity involves principled choices about where AI is integrated in teaching and learning sequences. It also involves clarity about the knowledge that students are expected to apply during ‘conversations’ with AI. In this regard, the paper draws attention to a social semiotic reading of Vygotsky's semiotic mediation. It argues that anticipating what is mediated is crucial for the process of customising a chatbot and making new knowledge visible to our students
Bayesian Power Spectral Density Estimation for LISA Noise Based on Penalized Splines With a Parametric Boost
No full textFlexible and accurate noise characterization is crucial for the precise estimation of gravitational-wave parameters. We introduce a Bayesian method for estimating the power spectral density (PSD) of long, stationary time series, explicitly tailored for Laser Interferometer Space Antenna (LISA) data analysis. Our approach models the PSD as the geometric mean of a parametric and a nonparametric component, combining the knowledge from parametric models with the flexibility to capture deviations from theoretical expectations. The nonparametric component is expressed by a mixture of penalized B splines. Adaptive, data-driven knot placement, performed once at initialization, removes the need for a reversible-jump Markov chain Monte Carlo, while hierarchical roughness-penalty priors prevent overfitting. Validation on simulated autoregressive (AR) data of order 4 [AR(4)] demonstrates estimator consistency and shows that well-matched parametric components reduce the integrated absolute error compared to an uninformative baseline, requiring fewer spline knots to achieve comparable accuracy. Applied to one year of simulated LISA -channel (univariate) noise, our method achieves relative integrated absolute errors of (10ˉ²), making it suitable for iterative analysis pipelines and multiyear mission data sets
The Presence and Impact of Autistic Child Comorbid Conditions and Their Relationship to Parent Well-being
No full textPurpose
The diagnosis of Autism Spectrum Disorder is rising globally, and as a long-life condition associated with high support needs, parents of Autistic children experience greater parenting stress and lower quality of life than parents raising typically developing children. However, while research has investigated the relationship between the severity of child autism symptoms and parenting stress, studies into comorbid conditions that likewise impair child function are not as common and often focus on a small subset of conditions. The aim of the current study was to estimate the frequency of the five most common comorbidities reported in the autism literature (Anxiety, ADHD, Intellectual Disability, Gastrointestinal Issues, Sleep Disorder) and relate them to parenting stress and health-related quality of life (HRQOL).
Methods
Using an internet-based survey, parent reports of their Autistic child’s comorbid conditions and the impact these have on their child’s function, parenting stress, and parental HRQOL were obtained from 453 parents residing in New Zealand. A global measure of parenting stress was obtained using the 18-item Parenting Stress Scale, while HRQOL ratings were obtained using the 36-Item Short Form Survey (SF-36).
Results
While many parents indicated the presence of comorbid conditions in their Autistic child, a substantial proportion were not formally diagnosed. A Linear Mixed-Effects Model indicated that child anxiety, Intellectual Disability, and ADHD had the greatest impact on both child and parent, however, subsequent multivariate analyses clarified that sleep disorder and Gastrointestinal Issues had the largest effect on parental stress and HRQOL, followed by ADHD. This result was robust irrespective of whether parents were asked if the comorbidity was present (vs. absent), diagnosed (vs. undiagnosed), or when related to child (i.e., functional) and parent (i.e., stress) impact.
Conclusion
Evidence that child sleep disorder, Gastrointestinal Issues, and ADHD are most detrimental to parental well-being indicate that interventions targeting these comorbidities should be prioritised. Coupled with increased child function as a direct result of intervention, better parental outcomes should increase child well-being and family quality of life, indicating that future research into the diagnostic barriers associated with comorbid conditions would be useful
Beyond the Generalist
No full textInternational frameworks and accreditations define the core competencies required of information technology (IT) project managers. Among these, technical skills are often cited as important, particularly in IT-focused projects. However, the technical competencies required—and the extent to which project managers should possess them—remain unclear. The literature on this topic is limited, though existing studies indicate that technical proficiency contributes to project success in technical domains. To explore this gap, semi-structured interviews with IT project managers and project participants were undertaken to examine perceptions of technical skills. Findings reveal a divide between participants with technical education, who emphasized the necessity of technical expertise, whereas those without technical qualifications highlighted communication, motivation, and attitude as most critical. The study contributes insights into the strategic value that technical capability adds to IT project management effectiveness through the strategic capability model for technical project management.</p
Ensuring Reliable District Heating Systems: Identifying Critical Components under Independent and Cascading Failure Scenarios
No full textUrban 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
CPR Training Needs Reviving in Aotearoa New Zealand [Letter]
No full textWe wholeheartedly support the view that life-saving cardiopulmonary resuscitation (CPR) training, including automated external defibrillator (AED) use, should not be limited to healthcare professionals and should be mandatory in the Aotearoa New Zealand school curriculum. Children aged 13–14 have been shown to perform chest compressions comparable to adults, and younger children can be taught basic skills, such as calling emergency services or instructing an adult how to perform CPR
Can Administrative Data Be Used for a National Register of Hospitalised Stroke Patients? A New Zealand Validation Study
No full textBackground: Using community-based incidence studies and clinical registries to assess stroke care and outcomes is resource intensive and often geographically limited. Linked administrative data are lower-cost and wider-reaching, but potentially less accurate and complete. This study compared administrative data to national hospital-based study data to assess whether administrative data represents a valid alternative. Methods: We linked and compared data from the REGIONS Care Study, a New Zealand nationwide observational study, with administrative data from Statistics New Zealand's Integrated Data Infrastructure (IDI). Sensitivity, specificity, positive predictive value, and Cohen's kappa coefficient were used to assess case identification, risk factors, post-stroke outcomes, and interventions as applicable. Additional audits explored the validity of IDI ‘true false positives.’ Findings: From May to July 2018, 1719 patients with stroke were captured in REGIONS Care and 1833 in the IDI. Using REGIONS Care as the reference standard, the sensitivity of the IDI for stroke case identification was 83% and the positive predictive value 77%. There were 300 false-negatives and 414 false positives. The audit of two hospitals showed that some cases identified in IDI but excluded by REGIONS were actual strokes. For stroke risk factors, the IDI showed high sensitivity and specificity for diabetes (93% and 91%, respectively), atrial fibrillation (87% and 90%), and smoking (71% and 97%) but lower specificity for hypertension (61%), and dyslipidaemia (52%). A derived IDI favourable outcome measure showed good agreement with the modified Rankin Scale (sensitivity 88%, specificity 82%, kappa 0.67). The IDI accurately identified post-stroke medication use (sensitivities 81%–94%, specificities 78%–91%) and thrombectomy interventions (sensitivity 88%, kappa 0.91). Interpretation: The use of administrative data to ascertain stroke cases, risk factors, interventions and outcomes was feasible and compared well with manual hospital data collection making an administrative data based national stroke register possible, although supplementary data collection for comprehensive care evaluation may be required. Funding: The study was funded by the NZ Health Research Council (HRC 17/037)