International Journal of Complexity in Applied Science and Technology (10 papers in press)

Regular Issues

• Machine Learning Models based on Financial Data for Stock Trend Predictions  
  by John Phan, Hung-Fu Chang 
  Abstract: This paper investigates the application of long short-term memory (LSTM), one-dimensional convolutional neural networks (1D CNN), and logistic regression (LR), for predicting stock trends based on fundamental analysis. This research emphasises a companys financial statements and its intrinsic value for stock price trend forecasting. Using a dataset of 269 data points from publicly traded companies across various sectors from 2019 to 2023, we employ key financial ratios and the Discounted Cash Flow (DCF) model for two tasks: Annual Stock Price Difference (ASPD) and Difference between Current Stock Price and Intrinsic Value (DCSPIV). Assessing the likelihood of profitability from relationship between financial data and price action, and the current discrepancy between true value and market price, respectively. Our results demonstrate that LR models outperform CNN and LSTM models, achieving an average test accuracy of 74.66% for ASPD and 72.85% for DCSPIV, highlighting the benefits for portfolio managers in their decision-making processes.
  Keywords: Stock Trend Prediction; Fundamental Analysis; Machine Learning; CNN; LSTM; Logistic Regression.
  DOI: 10.1504/IJCAST.2025.10069470
   
  
• Brain tumour identification using improved YOLOv8  
  by Rupesh Dulal, Rabin Dulal 
  Abstract: Accurately identifying the extent of brain tumours remains a major challenge in brain cancer treatment, primarily due to the difficulty in detecting tumour boundaries from MRI scans. Manual detection is time-consuming and requires expert knowledge. In this study, we propose a modified YOLOv8 model for precise brain tumour detection in MRI images. We replaced the traditional non-maximum suppression (NMS) with a real-time detection transformer (RT-DETR) to eliminate hand-designed filtering. Additionally, we integrated ghost convolution to reduce computational costs while maintaining accuracy, and introduced a vision transformer block in the backbone to enhance context-aware feature extraction. The model was trained and tested on a publicly available brain tumour dataset. Experimental results show that our modified YOLOv8 outperforms the original YOLOv8 and other popular object detectors including faster R-CNN, mask R-CNN, YOLOv3-v5, SSD, RetinaNet, EfficientDet, and DETR, achieving a mAP@0.5 of 0.91.
  Keywords: brain tumour detection; deep learning; attention; transformer; YOLOv8.
  DOI: 10.1504/IJCAST.2025.10071167
   
  
• FourierNAT: a Fourier-mixing-based non-autoregressive transformer for parallel sequence generation  
  by Andrew Kiruluta 
  Abstract: We present FourierNAT, a novel non-autoregressive transformer (NAT) architecture that leverages Fourier-based mixing in the decoder to generate output sequences in parallel. While traditional NAT approaches often face challenges in capturing global dependencies, our method uses a discrete Fourier transform with learned frequency-domain gating to mix token embeddings across the entire sequence dimension. This design enables efficient propagation of context without explicit autoregressive steps. Empirically, FourierNAT achieves competitive results on WMT14 En-De and CNN/DailyMail benchmarks, highlighting that frequency-domain operations can mitigate coherence gaps often associated with NAT generation. Our approach underscores the potential of integrating spectral-domain operations to accelerate and improve parallel text generation.
  Keywords: non-autoregressive transformer: NAT; Fourier mixing; parallel sequence generation; global spectral operations; NAT architecture.
  DOI: 10.1504/IJCAST.2025.10071491
   
  
• A comprehensive review of machine learning techniques for detecting fraud in banking and payment services  
  by Sushmita Kumari, Kamlesh Kumar, Ashutosh Gaurav 
  Abstract: In today’s digital age, fraud detection in financial services has become essential due to the rapid growth and complexity of online transactions. Machine learning techniques are widely used to detect unusual activities in real time. This paper focuses on fraud in banking and payment services and proposes a three-stage review framework: formulating research questions, defining the research methodology, and analysing existing literature. The review reveals that supervised and unsupervised learning algorithms, such as Naïve Bayes, K-nearest neighbours, deep learning, support vector machine, decision tree, artificial neural network, XGBoost, and AdaBoost, are commonly applied for fraud detection. These models are evaluated using metrics like precision, recall, and F-score. Ensemble methods that combine multiple algorithms are also shown to improve detection accuracy. Finally, the review highlights future research directions, especially the need to strengthen wallet payment systems by developing more robust and efficient fraud detection algorithms to ensure secure digital transactions.
  Keywords: fraud detection; mobile payment; machine learning; unsupervised learning; supervised learning.
  DOI: 10.1504/IJCAST.2025.10072040
   
  
• A comparative analysis of optimisation methods for classification on various datasets  
  by Simanta Das, Soumitra Das 
  Abstract: Optimisation studies how a variety of mathematical structures can be analysed through the minimisation or maximisation of a function. In deep learning (DL), optimisation encompasses everything from hyperparameter tuning to weight and bias adjustment until the convergence of a loss or cost function (J), such that the model’s performance metrics and reliability in classification and regression are increased. Over the last few years, stochastic gradient methods and their variants - or adaptive gradient methods - have become very popular, with varying levels of success or otherwise. This study provides a neat comparison of adaptive gradient methods with respect to their accuracies and cross-entropy loss (CEL) in the mentioned tasks; it tested nine optimisation algorithms across three CNN architectures on MNIST, Fashion-MNIST, and CIFAR-10 datasets over 30 epochs. The overall top-performing optimisers were SGD, RMSProp, Adam, and Nadam, whereas Adagrad and Adadelta consistently performed lower.
  Keywords: adaptive gradient methods; optimisation methods; convolutional neural networks; CNNs; MNIST; FashionMNIST; CIFAR10.
  DOI: 10.1504/IJCAST.2025.10072041
   
  
• PreStroke_ML: a machine learning approach to heat stroke prediction  
  by Md. Zahurul Haque, Mimuza Tazvia, Afreen Sultana Kuna 
  Abstract: Heatstroke is an increasingly critical public health concern, intensified by rising global temperatures and the growing frequency of extreme heat events. This study addresses the urgent need for timely and accurate heatstroke risk prediction by leveraging machine learning techniques. The primary objective is to develop a predictive model capable of identifying individuals at risk based on environmental and physiological data. An extensive dataset of 81,215 instances and 69 features underwent thorough preprocessing and analysis. Four machine learning algorithms - decision tree, random forest, logistic regression, and light gradient boosting machine (LightGBM) - were implemented and evaluated. Among these, LightGBM achieved the highest accuracy of 99.93%, demonstrating superior predictive performance and generalisation capability, as validated through confusion matrices and trainingvalidation accuracy curves. Feature selection played a crucial role in optimising model effectiveness. The findings underscore the potential of machine learning as a valuable tool in predictive healthcare. Future work will focus on integrating real-time sensor data, enabling personalised risk assessments, and deploying a mobile-based alert system to enhance heatstroke prevention. This research contributes to proactive public health strategies through an AI-driven framework for early detection and intervention.
  Keywords: heatstroke prediction; AI in healthcare; public health; early warning system; health risk prevention.
  DOI: 10.1504/IJCAST.2025.10072080
   
  
• A survey of identification and forecasting of healthcare fraud through machine learning  
  by Rasanarayan Chaurasiya, Kirti Jain, Vikas Chaurasia 
  Abstract: Healthcare fraud is a widespread problem that costs billions of dollars annually and has significant societal and financial consequences. Patients may face increased premiums and out-of-pocket expenses as a result of this because it compromises the integrity of healthcare systems. This survey analyses the ongoing philosophies for medical services misrepresentation recognition and expectation, the difficulties confronted, and arising patterns in this basic field. Patients and providers alike are harmed by healthcare fraud, which can lead to decreased quality and increased costs. The vast, complex, and ever-evolving nature of healthcare data has proven to be too much for traditional fraud detection methods. Improved fraud detection and prediction in the healthcare industry may be possible with the help of machine learning. This review looks at how various ML techniques are used to find healthcare fraud, talks about the problems and opportunities, and gives ideas for where research and practice should go in the future.
  Keywords: healthcare fraud; forecasting; machine learning; ML; fraud detection; premiums and expenses.
  DOI: 10.1504/IJCAST.2025.10072845
   
  
• Deep Learning-Driven Real-Time Violence Detection in Surveillance Streams  
  by Avi Vera 
  Abstract: The escalating threat of violence in public spaces necessitates scalable, automated, and real-time detection systems. This study introduces a deep learning-based framework for real-time violence detection in surveillance streams, leveraging a fine-tuned DenseNet121 convolutional neural network optimised for processing real-time streaming protocol (RTSP) feeds. Trained on a curated subset of the UCF-Crime dataset, the model achieves 92% accuracy and a weighted F1-score of 0.91. Integrating OpenCV for frame capture, flask for visualisation, MongoDB for metadata management, and Dropbox for cloud storage, the system processes multiple RTSP streams concurrently at 30 fps on a T4 GPU. This end-to-end pipeline offers a practical solution for smart city surveillance, transportation hubs, and institutional security, demonstrating scalability, robustness, and deployability. This manuscript extends our previous work previously shared as preprints to promote open science and reproducibility. It is available as a preprint on SSRN (Verma, 2025a), TechRxiv (Verma, 2025b) and on Zenodo (Verma, 2025c). The complete source code, model files, and deployment instructions for the proposed real-time violence detection system are available at GITHUB and dataset at DATASET.
  Keywords: Violence Detection; DenseNet121; Real-Time Surveillance; Deep Learning; Multithreading; Cloud Integration.
  DOI: 10.1504/IJCAST.2025.10073735
   
  
• Graph-Enhanced Retrieval-Augmented Question Answering for E-Commerce Customer Support  
  by Piyushkumar Patel 
  Abstract: E-commerce customer support requires quick and accurate answers grounded in product data and past support cases. This paper develops a novel retrieval-augmented generation (RAG) framework that uses knowledge graphs (KGs) to improve the relevance of the answer and the factual grounding. We examine recent advances in knowledge-augmented RAG and chatbots based on large language models (LLM) in customer support, including Microsofts GraphRAG and hybrid retrieval architectures. We then propose a new answer synthesis algorithm that combines structured subgraphs from a domain-specific KG with text documents retrieved from support archives, producing more coherent and grounded responses. We detail the architecture and knowledge flow of our system, provide comprehensive experimental evaluation, and justify its design in real-time support settings. Our implementation demonstrates 23% improvement in factual accuracy and 89% user satisfaction in e-commerce QA scenarios.
  Keywords: Retrieval-Augmented Generation; Knowledge Graph; Question Answering; Customer Support; E-Commerce; Large Language Models.
  DOI: 10.1504/IJCAST.2025.10073851
   
  
• SuperHyperGraph Neural Network and Dynamic SuperHyperGraph Neural Network  
  by Takaaki Fujita 
  Abstract: Graph theory explores the relationships between objects through mathematical structures composed of vertices (nodes) and edges (connections). A HyperGraph generalises the classical graph by introducing hyperedges, which can connect any number of vertices rather than just two, thus enabling the modelling of more complex multi-way relationships. Building upon this, the concept of a SuperHyperGraph has been introduced as a further extension of HyperGraphs and has recently become a subject of active research. Graph neural networks (GNNs) are one of the most extensively studied frameworks in artificial intelligence. The HyperGraph neural network (HGNN) extends GNNs by leveraging the expressive power of HyperGraphs. In this paper, we provide a concise introduction to the n-SuperHyperGraph Neural Network, which mathematically extends the HGNN architecture using SuperHyperGraphs. We also explore the concept of a dynamic n-SuperHyperGraph neural network, inspired by the ideas behind dynamic graph neural networks and dynamic HyperGraph neural networks. We anticipate that these formal developments will pave the way for future computational experiments on real-world datasets.
  Keywords: HyperGraph; SuperHyperGraph; Graph Theory; Graph Neural Networks; HyperGraph Neural Networks; SuperHyperGraph Neural Networks.
  DOI: 10.1504/IJCAST.2025.10073956