Natalija Stojanovic

Dr. Natalija Stojanovic

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 March 2026 in Future Internet
Verified
Journal Article
Open access
PDF file
AIAleksandra Stojnev Ilic
Aleksandra Stojnev Ilic
MIMilos Novica Ilic
Milos Novica Ilic
NSNatalija Stojanovic
Natalija Stojanovic
DSDragan Stojanović
Dragan Stojanović
Continuous healthcare monitoring systems generate non-stationary physiological data streams, where evolving statistical properties and patterns often invalidate static models and fixed user classifications. To address this challenge, we propose drift-aware adaptive architecture that integrates concept drift detection into a distributed edge–cloud data analytics pipeline. In the proposed design, a concept drift is elevated from a maintenance signal to the primary mechanism governing user-state adaptation, model evolution, and inference consistency. Within the proposed system, the edge tier performs low-latency inference and preliminary drift screening under strict resource constraints, while the cloud tier executes advanced drift detection and validation, orchestrates user reclassification and model retraining, and manages model evolution. A feedback loop synchronizes edge and cloud operations, ensuring that detected drift triggers appropriate system transitions, either reassigning a user to an updated state category or initiating targeted model updates. This architecture reduces reliance on static group assignments, improves personalization, and preserves model fidelity under evolving physiological conditions. We analyze the drift types most relevant to healthcare data streams, evaluate the suitability of lightweight and cloud-grade drift detectors, and define the system requirements for stability, responsiveness, and clinical safety. Evaluation across 21 concurrent users demonstrates that drift-aware adaptation reduced prediction MAE by 40.6% relative to periodic retraining, with an end-to-end adaptation latency of 66 ± 37 s. Hierarchical cloud validation reduced the false-positive retraining rate from 88.9% (edge-only triggering) to 27.3%, while maintaining uninterrupted inference throughout all adaptation events.

Verified
Journal Article
Open access
PDF file
NSNatalija Stojanovic
Natalija Stojanovic
DSDragan Stojanović
Dragan Stojanović
Watershed analysis, as a fundamental component of digital terrain analysis, is based on the Digital Elevation Model (DEM), which is a grid (raster) model of the Earth surface and topography. Watershed analysis consists of computationally and data intensive computing algorithms that need to be implemented by leveraging parallel and high-performance computing methods and techniques. In this paper, the Multiple Flow Direction (MFD) algorithm for watershed analysis is implemented and evaluated on multi-core Central Processing Units (CPU) and many-core Graphics Processing Units (GPU), which provides significant improvements in performance and energy usage. The implementation is based on NVIDIA CUDA (Compute Unified Device Architecture) implementation for GPU, as well as on OpenACC (Open ACCelerators), a parallel programming model, and a standard for parallel computing. Both phases of the MFD algorithm (i) iterative DEM preprocessing and (ii) iterative MFD algorithm, are parallelized and run over multi-core CPU and GPU. The evaluation of the proposed solutions is performed with respect to the execution time, energy consumption, and programming effort for algorithm parallelization for different sizes of input data. An experimental evaluation has shown not only the advantage of using OpenACC programming over CUDA programming in implementing the watershed analysis on a GPU in terms of performance, energy consumption, and programming effort, but also significant benefits in implementing it on the multi-core CPU.

8 December 2020 in Facta Universitatis, Series: Automatic Control and Robotics
Verified
Journal Article
Closed access
PDF file
NSNatalija Stojanovic
Natalija Stojanovic
DSDragan Stojanović
Dragan Stojanović
With the overpopulation of large cities, the problems with citizens’ mobility, transport inefficiency, traffic congestions and environmental pollution caused by the heavy traffic require advanced ITS solutions to be overcome. Recent advances and wide proliferation of mobile and Internet of Things (IoT) devices, carried by people, built in vehicles and integrated in a road infrastructure, enable collection of large scale data related to mobility and traffic in smart cities, still with a limited use in real world applications. In this paper, we propose the traffic monitoring, control and adaptation platform, named TrafficSense, based on Big Mobility Data processing and analytics. It provides a continuous monitoring of a traffic situation and detection of important traffic parameters, conditions and events, such as travel times along the street segments and traffic congestions in real time. Upon detecting a traffic congestion on an intersection, the TrafficSense application leverages the feedback control loop mechanism to provide a traffic adaptation based on the dynamic configuration of traffic lights duration in order to increase the traffic flows in critical directions at the intersections. We tested and evaluated the developed application on the distributed cloud computing infrastructure. By varying the streaming workload and the cluster parameters we show the feasibility and applicability of our approach and the platform.

3 June 2024
Verified
Conference Paper
Closed access
DSDragan Stojanović
Dragan Stojanović
DJDušan Jovanović
Dušan Jovanović
NSNatalija Stojanovic
Natalija Stojanovic
Big medical data analytics has become an essential research area in the healthcare industry. The vast amount of data generated by various medical devices, clinical systems, and research studies needs to be processed, analyzed and visualized to extract valuable insights. In this paper, we present methods and techniques for processing and analyzing Big medical data using Apache Spark, a widely used big data processing framework. We present three case studies using different types and formats of COVID-19 related data. We evaluate the performance of processing and analytics of these datasets using Apache Spark framework and a cluster of Spark Docker containers. Our experiments demonstrate the feasibility of using Apache Spark for Big medical data analytics.

13 October 2025
Verified
Conference Paper
Closed access
EĆEmilija Ćojbašić
Emilija Ćojbašić
DSDragan Stojanović
Dragan Stojanović
DRDejan Rančić
Dejan Rančić
NSNatalija Stojanovic
Natalija Stojanovic
ZCZarko Cojbasic
Zarko Cojbasic
The integration of Explainable Artificial Intelligence (XAI) and Edge Artificial Intelligence has been investigated to enhance transparency and trust in intelligent systems for industrial robotics. A real-time hand gesture recognition system has been implemented using the YOLO deep learning model on the resource-constrained NVIDIA Jetson Orin Nano platform. To address the need for interpretability, several XAI techniques-SHAP, LIME, and Grad-CAM-have been applied to analyze and visualize decision-making process of the model. Performance and resource utilization metrics have been collected for each XAI method to evaluate trade-offs between interpretability and system efficiency. The limitations imposed by the edge hardware have been identified as a critical factor affecting the feasibility of real-time explanations. Wider conclusions have been devised and future research directions outlined. The study demonstrated the practical viability of deploying interpretable AI models on edge devices, contributing toward more transparent and interactive human-robot collaboration in industrial settings.

1 January 2025 in Computer Science and Information Systems
Verified
Journal Article
Open access
PDF file
DSDragan Stojanović
Dragan Stojanović
SSStefan Sentic
Stefan Sentic
NSNatalija Stojanovic
Natalija Stojanovic
TSTeodora Stamenkovic
Teodora Stamenkovic
Edge computing and edge intelligence have gained significant traction in recent years due to the proliferation of Internet of Things devices, the exponential growth of data generated at the network edge, and the demand for real-time and context-aware applications. Despite its promising potential, the application of artificial intelligence on the edge faces many challenges, such as edge computing resource constraints, heterogeneity of edge devices, scalability issues, security and privacy concerns, etc. The paper addresses the challenges of deploying deep neural networks for edge intelligence and traffic object detection and recognition on a video captured by edge device cameras. The primary aim is to analyze resource consumption and achieve resource-awareness, optimizing computational resources across diverse edge devices within the edge-fog computing continuum while maintaining high object detection and recognition accuracy. To accomplish this goal, a methodology is proposed and implemented that exploits the edge-to-fog paradigm to distribute the inference workload across multiple tiers of the distributed system architecture. The edge-fog related solutions are implemented and evaluated in several use cases on datasets encompassing real-world traffic scenarios and traffic objects? recognition problems, revealing the feasibility of deploying deep neural networks for object recognition on resource-constrained edge devices. The proposed edge-to-fog methodology demonstrates enhancements in recognition accuracy and resource utilization, validating the viability of both edge-only and edge-fog based approaches. Furthermore, experimental results demonstrate the system?s adaptability to dynamic traffic scenarios, ensuring real-time recognition performance even in challenging environments.