International Journal of Complexity in Applied Science and Technology (14 papers in press)

Regular Issues

• 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
   
  
• 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
   
  
• Hybrid CNN with Chebyshev Polynomial Expansion for Medical Image Analysis  
  by Abhinav Roy, Bhavesh Gyanchandani, Aditya Oza 
  Abstract: Lung cancer remains a leading cause of cancer-related deaths, where early and accurate diagnosis is vital. Automated detection of pulmonary nodules in CT scans is challenging due to variations in nodule characteristics. While CNNs have shown promise, they struggle to capture fine-grained spatial-spectral features. We propose a hybrid CNN architecture enhanced with Chebyshev polynomial expansions, leveraging their orthogonality and approximation properties to extract high-frequency features and improve non-linear function modeling. Evaluated on LUNA16 and LIDC-IDRI datasets, our model outperforms standard CNNs in classifying nodules as benign or malignant, achieving notable gains in accuracy, sensitivity, and specificity. This method offers a robust framework for medical image analysis and clinical decision support.
  Keywords: Chebyshev Polynomials; Convolutional Neural Networks; Deep Learning; Function Approximation; Hybrid Deep Learning Model.
  DOI: 10.1504/IJCAST.2026.10074363
   
  
• A Comparative Study of Machine Learning and Grey Approach for Forecasting FX Rates  
  by Noorshanaaz Khodabaccus, Aslam A. E. F. Saib 
  Abstract: Volatility in the foreign exchange (FX) market is often associated with risk and can disrupt the sustainable development of an economy. The Mauritian economy, being open and globally integrated, is highly sensitive to currency fluctuations. Consequently, modelling and forecasting FX market volatility is crucial for proper risk management. This paper presents a comparative study on FX rate modelling and forecasting accuracy, contrasting conventional deep learning approaches with grey models. In particular, we compare the performance of recurrent neural networks (RNNs) and long short-term memory (LSTM) network,s using high-frequency historical data against the basic grey model and the optimised Fourier grey Markov model (FOGM), which relies on a significantly smaller dataset. Our findings indicate that the FOGM model, despite using a smaller dataset, outperforms the deep learning approaches considered.
  Keywords: FX rates forecasting; Volatility; Deep learning; Grey model; Optimised Fourier grey Markov model.
  DOI: 10.1504/IJCAST.2025.10074547
   
  
• Advancing Deep Learning Techniques for Low-Resource Shahmukhi Punjabi Language Processing  
  by Muhammad Shabbir, Mudassir Iftikhar 
  Abstract: Shahmukhi Punjabi, over 18 million of Pakistani speaks the Shahmukhi Punjabi language but there is not proper research is being available yet. So this paper explores the software of named entity recognition (NER), recurrent neural network (RNN), and long short-term memory (LSTM) models on a dedicated dataset. The study includes a thorough analysis of loss graphs, accuracy measures, and understanding matrices. Our contribution is looks into Shahmukhi Punjabis underappreciated work on a variety of sophisticated complex space models, including NER and network RNN. The majority of current research focuses on languages that are widely spoken. By implementing LSTM, RNN, and NER models and evaluating their efficiency on specific Shahmukhi Punjabi data, this work seeks to close this gap on Shahmukhi language research. Our LSTM models give us the 82% accuracy and RNN give us the 82.57%.
  Keywords: NLP; Deep Learning; RNN; LSTM; Data Science.
  DOI: 10.1504/IJCAST.2026.10075028
   
  
• A Model-Controller-Presenter Tri-Layer Control-Theoretic Orchestration Framework for Synergistic Efficiency and Interpretability in Multimodal Large Language Models  
  by Yaolin Zhang, Menghui Li 
  Abstract: Targeting the issues of insufficient computational efficiency and limited interpretability encountered by large-scale models when dealing with complex tasks such as multi-turn reasoning and multi-modal cooperation, this research puts forward a three-tier collaborative framework centred on model-controller-task adaptation (MCP). By decoupling the functionalities of large models into reasoning, generation, and retrieval modules, and integrating dynamic routing algorithms driven by reinforcement learning and task adaptation mechanisms, the systematic integration of control theory and the dynamic reasoning of large models is realised for the first time. Experimental results demonstrate that the MCP framework enhances the performance on cross-modal benchmark tasks (e.g., GLUE, COCO, ScienceQA) by 1530% compared with the baseline model, improves reasoning efficiency by 40%, and generates interpretable intermediate results through the presenter layer, achieving a manual interpretability score of 90%. This offers a brand-new technological route to tackle the bottleneck in the practical application of large-scale models.
  Keywords: large model optimisation; MCP framework; dynamic control flow; reinforcement learning routing;.
  DOI: 10.1504/IJCAST.2025.10075441
   
  
• AI-Driven Prediction Model for Antenatal and Postpartum Depression Among Bangladeshi Pregnant Mothers  
  by M.D. Zahurul Haque, Tasnim Binta Anowar, Sumaiya Jannat Samira 
  Abstract: Antenatal and postpartum depression (APD) are significant maternal mental health concerns, especially in lowresource settings like Bangladesh. This study introduces an AI-driven prediction model aimed at identifying the risk of APD among Bangladeshi mothers. Data were gathered from over 500 participants via Google Forms distributed through hospitals, online platforms, and community networks. The dataset encompasses a range of demographic, psychological, and lifestyle factors. Machine learning algorithms-random forest, XGBoost and gradient boosting were employed, demonstrating high accuracy in predicting depression severity. The developed web-based application enables real-time risk assessments, facilitating early detection and timely intervention. This research highlights the transformative role of AI in enhancing maternal mental health services and delivering scalable, data-driven solutions in resource-limited environments.
  Keywords: Antenatal depression; Postpartum depression; Machine learning; Mental health prediction; AI in healthcare.
  DOI: 10.1504/IJCAST.2025.10075720
   
  
• Comparative Study between a GA-Tuned PID Controller based on Minimising IAE, ITAE, and ISE Objective Functions in a Shell and Tube Heat Exchanger Temperature Control System  
  by Melat Getachew Kebede  
  Abstract: A heat exchanger device is widely used in process industries because it is capable of sustaining a wide range of temperatures. The heat exchanger temperature control system is a highly nonlinear, time-delayed, and complex system. additionally, it is accompanied by the presence of flow variation and temperature variation of input fluid. In this study, performance of a genetic algorithm-based proportional integral and derivative controller tuned based on minimising IAE, ITAE, and ISE objective functions has been analysed and compared under four scenarios, namely no disturbance, flow variation disturbance, temperature variation disturbance, and both disturbance conditions. Among the three objective functions used, the overall performance of the genetic algorithm-based proportional integral and derivative controller tuned based on minimising the time integral of the absolute error objective function is better than the one tuned based on minimizing the integral of the absolute error and the integral of the squared error objective functions.
  Keywords: : Genetic algorithm; GA tuned PID controller; Integral Square Error; Integral Absolute Error and Integral Time Absolute Error.
  DOI: 10.1504/IJCAST.2025.10076038
   
  
• A Comprehensive Analysis of Classical Sorting Algorithms with Diverse Input Conditions  
  by Mohsen Mohammadagha 
  Abstract: This study presents a comprehensive analysis of six classical sorting algorithms Mergesort, Heapsort, Quicksort (with median-of-three pivot), Insertion Sort, Selection Sort, and Bubble Sort to evaluate their practical efficiency across diverse input conditions. While theoretical complexity (O(n2) vs. O(nlog(n))) provides foundational insights, real-world performance depends on implementation-specific factors, input size, and data distribution. The research addresses the critical need to bridge theoretical predictions with empirical benchmarks, particularly as modern computing environments demand optimised algorithm selection for varying workloads. Using Python-based implementations, the methodology systematically tests algorithms on arrays (size 10100,000) with randomised, sorted, reverse-sorted, and custom patterns, measuring execution times and memory usage. Results reveal quadratic algorithms outperform O(nlog(n)) methods for small datasets (e.g., Selection Sort: 0.000004s at n = 10), while Quicksort dominates at scale (0.089s vs. Bubble Sorts 265.93s at n = 100,000). Logarithmic visualisations highlight exponential efficiency divergence, with O(nlog(n)) algorithms achieving 2,900
  Keywords: Modeling; Optimization; Hybridization; Sorting Algorithms; Comparative Study.
  DOI: 10.1504/IJCAST.2025.10076087
   
  
• Quantum AI-Driven Cloud Framework for Intelligent Urban Surveillance  
  by Ranjan Kumar Mandal  
  Abstract: Rapid urbanization has intensified challenges in public safety, efficient resource allocation, and effective governance. Smart city initiatives address these concerns through advanced technological frameworks, with cloud computing and artificial intelligence (AI) playing critical roles in modern surveillance solutions. However, existing systems face limitations in real-time processing and timely decision-making when handling large-scale data streams. This paper presents a Quantum AI-Integrated Cloud Surveillance Architecture designed for smart cities. The proposed framework leverages quantum-enhanced computational capabilities within cloud infrastructure to accelerate large-scale data processing and improve scalability. Quantum AI algorithms further enhance analytical performance, enabling real-time video analysis, anomaly detection, and predictive modelling. Experimental evaluations demonstrate the system's potential to deliver proactive threat mitigation and optimised resource management, thereby offering a robust foundation for next-generation urban surveillance.
  Keywords: Smart city; cloud computing; quantum computing; artificial intelligence; surveillance systems; real-time analytics; anomaly detection; predictive modelling.
  DOI: 10.1504/IJCAST.2026.10076619
   
  
• A Comparative Study of NLP Systems for Sentiment Polarity Classification across Different Domains and Genres  
  by Vincenzo Sammartino 
  Abstract: Sentiment analysis systems are now among the most widely used tools across various sectors: from politics to stock markets, from marketing to communication, from the sports domain to medical and natural sciences, and from social media analysis to consumer preference evaluation. This study presents a performance comparison of different methodologies, techniques, and applications developed in recent years. We describe a concrete implementation of two distinct Natural Language Processing (NLP) systems for the sentiment polarity classification of Italian tweets and Amazon reviews. Two different classification systems were developed: the first employs an approach based on the explicit representation of the texts' linguistic features, while the second uses an approach based on non-interpretable vectors (embeddings). Finally, a study was conducted to understand which features are most relevant for classification, and the underlying causes that influence the systems' performance in both in-domain and out-of-domain scenarios are highlighted.
  Keywords: Sentiment Analysis; Natural Language Processing; Machine Learning; Support Vector Machines; Domain Adaptation; Text Classification; Bag-of-Words; Word Embeddings; Feature Engineering; Italian; UGC.
  DOI: 10.1504/IJCAST.2026.10076620