Faculty of Design and Creative Technologies (Te Ara Auaha)
Permanent link for this community
The Faculty of Design and Creative Technologies - Te Ara Auaha is comprised of four schools: The School of Future Environments - Huri Te Ao, the School of Art and Design - Te Kura Toi a Hoahoa, the School of Communication Studies - Te Kura Whakapāho and the School of Engineering, Computer and Mathematical Sciences - Te Kura Mātai Pūhanga, Rorohiko, Pāngarau. It also has Institutes, Centres and Labs across the Arts and Sciences in a mix that blends the traditional and the new, praxis and theory.
Browse
Browsing Faculty of Design and Creative Technologies (Te Ara Auaha) by Subject "01 Mathematical Sciences"
Now showing 1 - 6 of 6
Results Per Page
Sort Options
- ItemAn Adaptive Deep Learning Neural Network Model to Enhance Machine-Learning-Based Classifiers for Intrusion Detection in Smart Grids(MDPI AG, 2023-06-02) Li, Xue Jun; Ma, Maode; Sun, YihanModern smart grids are built based on top of advanced computing and networking technologies, where condition monitoring relies on secure cyberphysical connectivity. Over the network infrastructure, transported data containing confidential information, must be protected as smart grids are vulnerable and subject to various cyberattacks. Various machine learning based classifiers were proposed for intrusion detection in smart grids. However, each of them has respective advantage and disadvantages. Aiming to improve the performance of existing machine learning based classifiers, this paper proposes an adaptive deep learning algorithm with a data pre-processing module, a neural network pre-training module and a classifier module, which work together classify intrusion data types using their high-dimensional data features. The proposed Adaptive Deep Learning (ADL) algorithm obtains the number of layers and the number of neurons per layer by determining the characteristic dimension of the network traffic. With transfer learning, the proposed ADL algorithm can extract the original data dimensions and obtain new abstract features. By combining deep learning models with traditional machine learning-based classification models, the performance of classification of network traffic data is significantly improved. By using the Network Security Laboratory-Knowledge Discovery in Databases (NSL-KDD) dataset, experimental results show that the proposed ADL algorithm improves the effectiveness of existing intrusion detection methods and reduces the training time, indicating a promising candidate to enhance network security in smart grids.
- ItemAn Attention-Based CNN-BiLSTM Model for Depression Detection on Social Media Text(Elsevier BV, 2024-03-22) Philip Thekkekara, Joel; Yongchareon, Sira; Liesaputra, VeronicaDepression has long been described as a common mental health disorder and a disease with a set of diagnostic criteria that influences the affected individuals' feelings and behavior. The prevalence of Internet use has augmented people’s openness to share their experiences and struggles, including mental health disorders on social media thus researchers have tried developing classification models for depression detection using various machine learning and deep learning techniques. In this research, we propose a deep learning architecture with an attention mechanism on CNN-BiLSTM (CBA) and provide a comparative analysis to benchmark well-known deep learning models using the public dataset namely CLEF2017. We found that along with F1 score, precision and recall it is also vital to consider the Area under the curve - Receiver operating characteristic curve (AUC-ROC) and Mathews Correlation Coefficient (MCC) metrics for evaluating depression classification models since the MCC considers all the four values of a confusion matrix. Based on our experiments, the CBA model outperforms the existing state of the art model with an overall accuracy of 96.71% and scores of 0.85 and 0.77 for AUC-ROC and MCC, respectively.
- ItemElectron Beam Powder Bed Fusion Additive Manufacturing of Ti6Al4V Alloy Lattice Structures: Orientation-Dependent Compressive Strength and Fracture Behavior(Springer, 2024-04-09) Huang, Yawen; Chen, ZW; Wan, ARO; Schmidt, K; Sefont, P; Singamneni, SHigh porosity level lattice structures made using electron beam powder bed fusion additive manufacturing (EBPBF) need to be sufficiently strong and the understanding of the mechanical anisotropy of the structures is important for the design of orthopedic implants. In this work, the combined effects of loading direction (LD), cell orientation, and strut irregularity associated with EBPBF of Ti6Al4V alloy lattices on the mechanical behavior of the lattices under compressive loading have been studied. Three groups of simple cubic unit cell lattices were EBPBF made, compressively tested, and examined. The three groups were [001]//LD lattices, [011]//LD lattices, and [111]//LD lattices. Simulation has also been conducted. Yield strength (σy-L) values of all lattices determined experimentally have been found to be comparable to the values predicted by simulation; thus, EBPBF surface defects do not affect σy-L. σy-L of [001]//LD lattices is 1.8–2.0 times higher than those of [011]//LD and [111]//LD lattices. The reason for this is shown to be due to the high stress concentrations in non-[001]//LD samples, causing yielding at low loading levels. Furthermore, plastic strain (εp) at ultimate compression strength of [001]//LD samples has been determined to be 4–6 times higher than the values of non-[001]//LD samples. Examining the tested samples has shown cracks more readily propagating from EBPBF micro-notches in non-[001]//LD samples, resulting in low εp.
- ItemFurnace Vestibule Heat Transport Models(Australian Mathematical Publishing Association, Inc., ) McGuinness, Mark; Cox, Barry; Kalyanaraman, Balaje; Kiradjiev, Kristian; Gonzalez-Farina, Raquel; Hassell Sweatman, Catherine; Roberts, Lindon; Pontin, David; Bissaker, Edward; Irvine, Samuel; Jenkins, David; Taggart, IanThis is a report on the Lovells Springs challenge that was brought to the Mathematics in Industry Study Group at the University of Newcastle, Australia, in January 2020. The design of a furnace that heats steel rods to make them malleable and allow the reshaping of the rods into coiled springs is the challenge. Mathematical modelling of heat transport in the half-metre long furnace vestibule predicts the effect of vestibule geometry on the temperature of rods entering the furnace, and provides guidelines for deciding on the dimensions of the vestibule for improved energy efficiency of heating. Models considered include treating the rods as equivalent steel sheets, and as discrete steel rods. The relative importance of radiative and convective heat transfer mechanisms is considered. A longer vestibule, with length one or two metres, is recommended for improved heating efficiency of rods thicker than 25mm.
- ItemModelling Remission from Overweight Type 2 Diabetes Reveals How Altering Advice May Counter Relapse(Elsevier, 2024-03-20) Hassell Sweatman, CatherineThe development or remission of diet-induced overweight type 2 diabetes involves many biological changes which occur over very different timescales. Remission, defined by, or fasting plasma glucose concentration mg/dl, may be achieved rapidly by following weight loss guidelines. However, remission is often short-term, followed by relapse. Mathematical modelling provides a way of investigating a typical situation, in which patients are advised to lose weight and then maintain fat mass, a slow variable. Remission followed by relapse, in a modelling sense, is equivalent to changing from a remission trajectory with steady state mg/dl, to a relapse trajectory with steady state mg/dl. Modelling predicts that a trajectory which maintains weight will be a relapse trajectory, if the fat mass chosen is too high, the threshold being dependent on the lipid to carbohydrate ratio of the diet. Modelling takes into account the effects of hepatic and pancreatic lipid on hepatic insulin sensitivity and -cell function, respectively. This study leads to the suggestion that type 2 diabetes remission guidelines be given in terms of model parameters, not variables; that is, the patient should adhere to a given nutrition and exercise plan, rather than achieve a certain subset of variable values. The model predicts that calorie restriction, not weight loss, initiates remission from type 2 diabetes; and that advice of the form ‘adhere to the diet and exercise plan’ rather than ‘achieve a certain weight loss’ may help counter relapse.
- ItemPOI Recommendation for Occasional Groups Based on Hybrid Graph Neural Networks(Elsevier BV, 2023-09-19) Meng, L; Liu, Z; Chu, D; Sheng, QZ; Yu, J; Song, XRecently, POI (Point-of-interest) recommendation for groups has become a critical challenge when helping groups to discover potentially interesting new places, and some effective recommendation models have been proposed to address this issue. However, most existing research focuses on POI recommendation for fixed groups, few studies have been conducted on POI recommendation for occasional groups. To tackle this issue, we propose a POI recommendation model for occasional groups based on Hybrid Graph Neural Networks (termed as PROG-HGNN) which combines excellent graph neural networks models. Firstly, PROG-HGNN generates the fitted representation of the occasional group based on the Node Influence Indicator (INF) method and Graph Attention Networks (GAT) model. Then, PROG-HGNN learns POIs’ representations containing members’ POI interaction preferences and members’ POI transfer preferences with the Signed Bipartite Graph Neural Networks (SBGNN) model and the Session-based Graph Neural Networks (SRGNN) model, respectively. Finally, PROG-HGNN recommends the potential POIs for the occasional group based on the fitted representation of the occasional group and the learned representations of POIs. We verify our proposed model on three public benchmark datasets (Foursquare, Gowalla and Yelp), which contains 124,933 to 860,888 POI check-in records. The comparison between our proposed model and the twelve baseline models demonstrates the outstanding performance of PROG-HGNN. In terms of Precision@K and Recall@K, our model achieves about 32.92% and 19.67% improvement compared with the best baseline models on the three benchmark datasets averagely. Adequate ablation experiments prove the effectiveness of the members’ POI interaction preferences learning module and POI transfer preferences learning module.