Sazia Parvin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sazia Parvin has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Sitalakshmi Venkatraman
Department of Business and Construction, Melbourne Polytechnic, Preston, VIC 3072, Australia
7 shared publications - Kiran Fahd
Department of Business and Construction, Melbourne Polytechnic, Preston, VIC 3072, Australia
6 shared publications Joanna Jackson
Melbourne Polytechnic
3 shared publicationsAmjad Gawanmeh
University of Dubai
2 shared publications- Amjad Gawanmeh
College of Engineering and IT, University of Dubai, Dubai, UAE
2 shared publications Kiran Fahd
Melbourne Polytechnic
2 shared publications
Latest Publications
- Open Access
Why NoSQL Benchmarks Fail in AI Era: A Practical Validity-Aware Workflow for Trustworthy Data Store Evaluation
Selecting a data-management subsystem is a consequential decision in data-intensive applications, including AI pipelines that rely on low-latency storage. Such decisions are often informed by lightweight, application-specific benchmarks whose measurement validity is rarely examined. This paper presents a validity-aware benchmarking workflow for comparing NoSQL data stores under application-representative CRUD workloads. The workflow incorporates a final validation stage that checks whether measured operations exhibit the expected behaviour before performance comparisons are made. The approach is demonstrated using a key-value store (Aerospike) and a document store (MongoDB). The validity checks identified two operations that performed no actual database work. MongoDB’s idiomatic read operation returned a cursor that was never iterated, causing reads to appear approximately 130 times faster for single-record workloads and 1000 times faster for bulk workloads than equivalent Aerospike reads. In addition, a MongoDB bulk update matched no documents because update keys were constructed incorrectly. Further validity threats included a missing secondary index, unequal bulk workloads, and a first-run warmup effect. Consequently, the remaining write measurements are reported descriptively, together with confidence intervals and run-level results. The study does not establish which engine delivers superior performance. Instead, it demonstrates how seemingly reasonable application-specific benchmarks can generate misleading conclusions and argues for an explicit validity-assessment stage, supported by a threat model, before benchmark results are used to inform subsystem selection and architectural decision-making.
Open AccessWhy NoSQL Benchmarks Fail in AI Era: A Practical Validity-Aware Workflow for Trustworthy Data Store Evaluation
Selecting a data-management subsystem is a consequential decision in data-intensive applications, including AI pipelines that rely on low-latency storage. Such decisions are often informed by lightweight, application-specific benchmarks whose measurement validity is rarely examined. This paper presents a validity-aware benchmarking workflow for comparing NoSQL data stores under application-representative CRUD workloads. The workflow incorporates a final validation stage that checks whether measured operations exhibit the expected behaviour before performance comparisons are made. The approach is demonstrated using a key-value store (Aerospike) and a document store (MongoDB). The validity checks identified two operations that performed no actual database work. MongoDB’s idiomatic read operation returned a cursor that was never iterated, causing reads to appear approximately 130 times faster for single-record workloads and 1000 times faster for bulk workloads than equivalent Aerospike reads. In addition, a MongoDB bulk update matched no documents because update keys were constructed incorrectly. Further validity threats included a missing secondary index, unequal bulk workloads, and a first-run warmup effect. Consequently, the remaining write measurements are reported descriptively, together with confidence intervals and run-level results. The study does not establish which engine delivers superior performance. Instead, it demonstrates how seemingly reasonable application-specific benchmarks can generate misleading conclusions and argues for an explicit validity-assessment stage, supported by a threat model, before benchmark results are used to inform subsystem selection and architectural decision-making.
- Open Access
- 22 September 2026
- Published by GR Scholastic LLP
- Vol. 1 (2)
Q-MedDict: Quantum Dictionary Multimodality for Lung Tissue and Lab Data
A lung cancer record can contain very different kinds of data. A tissue slide shows cells and tissue shape. A blood test gives numbers such as white blood cell count, haemoglobin, albumin, glucose, and creatinine. A urine test gives numbers such as urine albumin and urine creatinine. Many computer models use only one of these data types. We propose Q-MedDict, a quantum-inspired dictionary model for image and laboratory data. The model stores information as key–value pairs. The key is a bit string. The value is an amplitude. Gate-like rules move or mix the amplitudes. The method follows the idea of dictionary-based quantum image encoding, but it adds a second branch for blood and urine data. The image branch was tested on two public lung adenocarcinoma H&E images. The laboratory branch used public NHANES release statistics for blood and urine variables. We did not pair CMB-LCA patients with NHANES people because they are different cohorts. The two branches were compared by the same encoding measures instead. With only 25% of block coefficients, the two lung images kept 99.80% and 99.35% of transform energy. Their SSIM values were 0.919 and 0.910. A 50% dictionary state for blood kept 98.64% of transform energy with cosine similarity 0.993. A 50% urine state kept 98.53% with cosine similarity 0.993. These results show one common pattern: image, blood, and urine information can be placed in the same dictionary form and reduced while most of the encoded signal stays present. The result is a small proof-of concept. It is not a lung cancer diagnostic model. A matched cohort with slides and laboratory tests from the same people is needed for clinical prediction.
- 22 September 2026
- Published by GR Scholastic LLP
- Vol. 1 (2)
Open AccessQ-MedDict: Quantum Dictionary Multimodality for Lung Tissue and Lab Data
A lung cancer record can contain very different kinds of data. A tissue slide shows cells and tissue shape. A blood test gives numbers such as white blood cell count, haemoglobin, albumin, glucose, and creatinine. A urine test gives numbers such as urine albumin and urine creatinine. Many computer models use only one of these data types. We propose Q-MedDict, a quantum-inspired dictionary model for image and laboratory data. The model stores information as key–value pairs. The key is a bit string. The value is an amplitude. Gate-like rules move or mix the amplitudes. The method follows the idea of dictionary-based quantum image encoding, but it adds a second branch for blood and urine data. The image branch was tested on two public lung adenocarcinoma H&E images. The laboratory branch used public NHANES release statistics for blood and urine variables. We did not pair CMB-LCA patients with NHANES people because they are different cohorts. The two branches were compared by the same encoding measures instead. With only 25% of block coefficients, the two lung images kept 99.80% and 99.35% of transform energy. Their SSIM values were 0.919 and 0.910. A 50% dictionary state for blood kept 98.64% of transform energy with cosine similarity 0.993. A 50% urine state kept 98.53% with cosine similarity 0.993. These results show one common pattern: image, blood, and urine information can be placed in the same dictionary form and reduced while most of the encoded signal stays present. The result is a small proof-of concept. It is not a lung cancer diagnostic model. A matched cohort with slides and laboratory tests from the same people is needed for clinical prediction.
- Open Access
Secure Fog Computing for Remote Health Monitoring with Data Prioritisation and AI-Based Anomaly Detection
Smart remote health monitoring requires time-critical medical data of patients from IoT-enabled cyber–physical systems (CPSs) to be securely transmitted and analysed in real time for early interventions and personalised patient care. Existing cloud architectures are insufficient for smart health systems due to their inherent issues with latency, bandwidth, and privacy. Fog architectures using data storage closer to edge devices introduce challenges in data management, security, and privacy for effective monitoring of a patient’s sensitive and critical health data. These gaps found in the literature form the main research focus of this study. As an initial modest step to advance research further, we propose an innovative fog-based framework which is the first of its kind to integrate secure communication with intelligent data prioritisation (IDP) integrated into an AI-based enhanced Random Forest anomaly and threat detection model. Our experimental study to validate our model involves a simulated smart healthcare scenario with synthesised health data streams from distributed wearable devices. Features such as heart rate, SpO2, and breathing rate are dynamically prioritised using AI strategies and rule-based thresholds so that urgent health anomalies are transmitted securely in real time to support clinicians and medical experts for personalised early interventions. We establish a successful proof-of-concept implementation of our framework by achieving high predictive performance measures with an initial high score of 93.5% accuracy, 90.8% precision, 88.7% recall, and 89.7% F1-score.
Open AccessSecure Fog Computing for Remote Health Monitoring with Data Prioritisation and AI-Based Anomaly Detection
Smart remote health monitoring requires time-critical medical data of patients from IoT-enabled cyber–physical systems (CPSs) to be securely transmitted and analysed in real time for early interventions and personalised patient care. Existing cloud architectures are insufficient for smart health systems due to their inherent issues with latency, bandwidth, and privacy. Fog architectures using data storage closer to edge devices introduce challenges in data management, security, and privacy for effective monitoring of a patient’s sensitive and critical health data. These gaps found in the literature form the main research focus of this study. As an initial modest step to advance research further, we propose an innovative fog-based framework which is the first of its kind to integrate secure communication with intelligent data prioritisation (IDP) integrated into an AI-based enhanced Random Forest anomaly and threat detection model. Our experimental study to validate our model involves a simulated smart healthcare scenario with synthesised health data streams from distributed wearable devices. Features such as heart rate, SpO2, and breathing rate are dynamically prioritised using AI strategies and rule-based thresholds so that urgent health anomalies are transmitted securely in real time to support clinicians and medical experts for personalised early interventions. We establish a successful proof-of-concept implementation of our framework by achieving high predictive performance measures with an initial high score of 93.5% accuracy, 90.8% precision, 88.7% recall, and 89.7% F1-score.
- Open Access
- 16 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Introducing Digital Twin Capability Building in Healthcare through AI Powered Projects
- 16 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Open AccessIntroducing Digital Twin Capability Building in Healthcare through AI Powered Projects
- Open Access
- 16 August 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (16) , 9047
A Set Theoretic Framework for Unsupervised Preprocessing and Power Consumption Optimisation in IoT-Enabled Healthcare Systems for Smart Cities
The emergence of the Internet of Things (IoT) has brought about a significant technological shift, coupled with the rise of intelligent computing. IoT integrates various digital and analogue devices with the Internet, enabling advanced communication between devices and humans.The pervasive adoption of IoT has transformed urban infrastructures into interconnected smart cities. Here, we propose a framework that mathematically models and automates power consumption management for IoT devices in smart city environments ranging from residential buildings to healthcare settings. The proposed framework utilises set theoretic association-rule mining and combines unsupervised preprocessing with frequent-item set mining and iterative numerical optimisation to reduce non-critical energy consumption. Readings are first converted into binary transaction matrices; then a modified Apriori algorithm is applied to extract high-confidence usage patterns and association rules. Dimensionality reduction techniques compress these transaction profiles, while the Gauss–Seidel method computes control set points that balance energy efficiency. The resulting rule set is deployed through a web portal that provides real-time device status, remote actuation, and automated billing. These associative rules generate predictive control functions, optimise the response of the framework, and prepare the framework for future events. A web portal is introduced that enables remote control of IoT devices and facilitates power usage monitoring, as well as automated billing.
- 16 August 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (16) , 9047
Open AccessA Set Theoretic Framework for Unsupervised Preprocessing and Power Consumption Optimisation in IoT-Enabled Healthcare Systems for Smart Cities
The emergence of the Internet of Things (IoT) has brought about a significant technological shift, coupled with the rise of intelligent computing. IoT integrates various digital and analogue devices with the Internet, enabling advanced communication between devices and humans.The pervasive adoption of IoT has transformed urban infrastructures into interconnected smart cities. Here, we propose a framework that mathematically models and automates power consumption management for IoT devices in smart city environments ranging from residential buildings to healthcare settings. The proposed framework utilises set theoretic association-rule mining and combines unsupervised preprocessing with frequent-item set mining and iterative numerical optimisation to reduce non-critical energy consumption. Readings are first converted into binary transaction matrices; then a modified Apriori algorithm is applied to extract high-confidence usage patterns and association rules. Dimensionality reduction techniques compress these transaction profiles, while the Gauss–Seidel method computes control set points that balance energy efficiency. The resulting rule set is deployed through a web portal that provides real-time device status, remote actuation, and automated billing. These associative rules generate predictive control functions, optimise the response of the framework, and prepare the framework for future events. A web portal is introduced that enables remote control of IoT devices and facilitates power usage monitoring, as well as automated billing.
- Open Access
- 1 February 2024
- Published by Springer Nature in Visual Computing for Industry, Biomedicine, and Art
- Vol. 7 (1) , 1-12
Convolutional neural network based data interpretable framework for Alzheimer’s treatment planning
Alzheimer’s disease (AD) is a neurological disorder that predominantly affects the brain. In the coming years, it is expected to spread rapidly, with limited progress in diagnostic techniques. Various machine learning (ML) and artificial intelligence (AI) algorithms have been employed to detect AD using single-modality data. However, recent developments in ML have enabled the application of these methods to multiple data sources and input modalities for AD prediction. In this study, we developed a framework that utilizes multimodal data (tabular data, magnetic resonance imaging (MRI) images, and genetic information) to classify AD. As part of the pre-processing phase, we generated a knowledge graph from the tabular data and MRI images. We employed graph neural networks for knowledge graph creation, and region-based convolutional neural network approach for image-to-knowledge graph generation. Additionally, we integrated various explainable AI (XAI) techniques to interpret and elucidate the prediction outcomes derived from multimodal data. Layer-wise relevance propagation was used to explain the layer-wise outcomes in the MRI images. We also incorporated submodular pick local interpretable model-agnostic explanations to interpret the decision-making process based on the tabular data provided. Genetic expression values play a crucial role in AD analysis. We used a graphical gene tree to identify genes associated with the disease. Moreover, a dashboard was designed to display XAI outcomes, enabling experts and medical professionals to easily comprehend the prediction results.
- 1 February 2024
- Published by Springer Nature in Visual Computing for Industry, Biomedicine, and Art
- Vol. 7 (1) , 1-12
Open AccessConvolutional neural network based data interpretable framework for Alzheimer’s treatment planning
Alzheimer’s disease (AD) is a neurological disorder that predominantly affects the brain. In the coming years, it is expected to spread rapidly, with limited progress in diagnostic techniques. Various machine learning (ML) and artificial intelligence (AI) algorithms have been employed to detect AD using single-modality data. However, recent developments in ML have enabled the application of these methods to multiple data sources and input modalities for AD prediction. In this study, we developed a framework that utilizes multimodal data (tabular data, magnetic resonance imaging (MRI) images, and genetic information) to classify AD. As part of the pre-processing phase, we generated a knowledge graph from the tabular data and MRI images. We employed graph neural networks for knowledge graph creation, and region-based convolutional neural network approach for image-to-knowledge graph generation. Additionally, we integrated various explainable AI (XAI) techniques to interpret and elucidate the prediction outcomes derived from multimodal data. Layer-wise relevance propagation was used to explain the layer-wise outcomes in the MRI images. We also incorporated submodular pick local interpretable model-agnostic explanations to interpret the decision-making process based on the tabular data provided. Genetic expression values play a crucial role in AD analysis. We used a graphical gene tree to identify genes associated with the disease. Moreover, a dashboard was designed to display XAI outcomes, enabling experts and medical professionals to easily comprehend the prediction results.
- 29 December 2023
- Published by IGI Global Scientific Publishing
- p. 1-20
A Biological Data-Driven Mining Technique by Using Hybrid Classifiers With Rough Set
Biological data classification and analysis are significant for living organs. A biological data classification is an approach that classifies the organs into a particular group based on their features and characteristics. The objective of this paper is to establish a hybrid approach with naive Bayes, apriori algorithm, and KNN classifier that generates optimal classification rules for finding biological pattern matching. The authors create combined association rules by using naïve Bayes and apriori approach with a rough set for next sequence prediction. First, the large DNA sequence is reduced by using k-nearest approach. They apply association rules by using naïve Bayes and apriori approach for the next sequence pattern. The hybrid approach provides more accuracy than single classifier for biological sequence prediction. The optimized hybrid process needs less execution time for rule generation for massive biological data analysis. The results established that the hybrid approach generally outperforms the other association rule generation approach.
- 29 December 2023
- Published by IGI Global Scientific Publishing
- p. 1-20
A Biological Data-Driven Mining Technique by Using Hybrid Classifiers With Rough Set
Biological data classification and analysis are significant for living organs. A biological data classification is an approach that classifies the organs into a particular group based on their features and characteristics. The objective of this paper is to establish a hybrid approach with naive Bayes, apriori algorithm, and KNN classifier that generates optimal classification rules for finding biological pattern matching. The authors create combined association rules by using naïve Bayes and apriori approach with a rough set for next sequence prediction. First, the large DNA sequence is reduced by using k-nearest approach. They apply association rules by using naïve Bayes and apriori approach for the next sequence pattern. The hybrid approach provides more accuracy than single classifier for biological sequence prediction. The optimized hybrid process needs less execution time for rule generation for massive biological data analysis. The results established that the hybrid approach generally outperforms the other association rule generation approach.
- Open Access
- 1 August 2023
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 13 (4) , 4306-4316
Big data analytics and internet of things for personalised healthcare: opportunities and challenges
With the increasing use of technologies and digitally driven healthcare systems worldwide, there will be several opportunities for the use of big data in personalized healthcare. In addition, With the advancements and availability of internet of things (IoT) based point-of-care (POC) technologies, big data analytics and artificial intelligence (AI) can provide useful methods and solutions in monitoring, diagnosis, and self-management of health issues for a better personalized healthcare. In this paper, we identify the current personalized healthcare trends and challenges. Then, propose an architecture to support big data analytics using POC test results of an individual. The proposed architecture can facilitate an integrated and self-managed healthcare as well as remote patient care by adapting three popular machine learning algorithms to leverage the current trends in IoT, big data infrastructures and data analytics for advancing personalized healthcare of the future.
- 1 August 2023
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 13 (4) , 4306-4316
Open AccessBig data analytics and internet of things for personalised healthcare: opportunities and challenges
With the increasing use of technologies and digitally driven healthcare systems worldwide, there will be several opportunities for the use of big data in personalized healthcare. In addition, With the advancements and availability of internet of things (IoT) based point-of-care (POC) technologies, big data analytics and artificial intelligence (AI) can provide useful methods and solutions in monitoring, diagnosis, and self-management of health issues for a better personalized healthcare. In this paper, we identify the current personalized healthcare trends and challenges. Then, propose an architecture to support big data analytics using POC test results of an individual. The proposed architecture can facilitate an integrated and self-managed healthcare as well as remote patient care by adapting three popular machine learning algorithms to leverage the current trends in IoT, big data infrastructures and data analytics for advancing personalized healthcare of the future.
- Open Access
- 12 December 2022
- Published by Research and Education Development (READ) Institute in International Journal of Asian Education
- Vol. 3 (4) , 211-224
Assessment Re-Think
Assessments are the fundamental media between students and educators. This paper aims to evaluate how to create assessments, how students learn from them, and how to link them to the industry and entrepreneurism. The implementation plan postulates how students can generate income from income-generation assessments or business innovation assessments. In this paper, we discuss the involvement of modern industry in assessment. We examine evidence from approximately 100 assessments detailed in 32 subject outlines. We employ a descriptive, pragmatic research methodology to consider whether they can be aligned more with industry expectations and expected duties. We propose a framework to connect with industry and create student income-generating projects. This proposed income-generating assessments framework recommended industry-based assessments with which students can not only earn marks towards a subject but potentially earn an income based on it. This paper extends the idea of peer learning to expert or industry learning: an approach that did not employ in higher education. Our approach supports educators in keeping the assessment up-to-date, enabling students to add more value to their learning of industry products and procedures. Students can directly contribute to the product and procedures and learn from the strategies actively employed in the workplace.
- 12 December 2022
- Published by Research and Education Development (READ) Institute in International Journal of Asian Education
- Vol. 3 (4) , 211-224
Open AccessAssessment Re-Think
Assessments are the fundamental media between students and educators. This paper aims to evaluate how to create assessments, how students learn from them, and how to link them to the industry and entrepreneurism. The implementation plan postulates how students can generate income from income-generation assessments or business innovation assessments. In this paper, we discuss the involvement of modern industry in assessment. We examine evidence from approximately 100 assessments detailed in 32 subject outlines. We employ a descriptive, pragmatic research methodology to consider whether they can be aligned more with industry expectations and expected duties. We propose a framework to connect with industry and create student income-generating projects. This proposed income-generating assessments framework recommended industry-based assessments with which students can not only earn marks towards a subject but potentially earn an income based on it. This paper extends the idea of peer learning to expert or industry learning: an approach that did not employ in higher education. Our approach supports educators in keeping the assessment up-to-date, enabling students to add more value to their learning of industry products and procedures. Students can directly contribute to the product and procedures and learn from the strategies actively employed in the workplace.
- Open Access
Developing an IoT Identity Management System Using Blockchain
Identity (ID) management systems have evolved based on traditional data modelling and authentication protocols that are facing security, privacy, and trust challenges with the growth of Internet of Things (IoT). Research surveys reveal that blockchain technology offers special features of self-sovereign identity and cryptography that can be leveraged to address the issues of security breach and privacy leaks prevalent in existing ID management systems. Although research studies are recently exploring the suitability of blockchain based support to existing infrastructure, there is a lack of focus on IoT ecosystem in the secured ID management with data provenance of digital assets in businesses. In this paper, we propose a blockchain based ID management system for computing assets in an IoT ecosystem comprising of devices, software, users, and data operations. We design and develop a proof-of-concept prototype using a federated and distributed blockchain platform with smart contracts to support highly trusted data storage and secure authentication of IoT resources and operations within a business case scenario.
Open AccessDeveloping an IoT Identity Management System Using Blockchain
Identity (ID) management systems have evolved based on traditional data modelling and authentication protocols that are facing security, privacy, and trust challenges with the growth of Internet of Things (IoT). Research surveys reveal that blockchain technology offers special features of self-sovereign identity and cryptography that can be leveraged to address the issues of security breach and privacy leaks prevalent in existing ID management systems. Although research studies are recently exploring the suitability of blockchain based support to existing infrastructure, there is a lack of focus on IoT ecosystem in the secured ID management with data provenance of digital assets in businesses. In this paper, we propose a blockchain based ID management system for computing assets in an IoT ecosystem comprising of devices, software, users, and data operations. We design and develop a proof-of-concept prototype using a federated and distributed blockchain platform with smart contracts to support highly trusted data storage and secure authentication of IoT resources and operations within a business case scenario.