← Back to Publications List

Comparative Analysis of Heap Algorithms for Anomaly Localization in Dynamic Graph Neural Networks

Students & Supervisors

Student Authors
Kiyas Mahmud
Bachelor of Science in Computer Science & Engineering, FST
Tauhid Sarker
Bachelor of Science in Computer Science & Engineering, FST
Syeda Shakira Akter
Bachelor of Science in Computer Science & Engineering, FST
Md. Maruf Hossain Munna
Bachelor of Science in Computer Science & Engineering, FST
Kazi Redwan
Bachelor of Science in Computer Science & Engineering, FST
Supervisors
Md. Faruk Abdullah Al Sohan
Lecturer, Faculty, FST

Abstract

Anomaly localization in dynamic graph neural networks (DGNNs) is very important to detect faults and abnormalities in complex systems like water supply and electricity networks. Online anomaly detection is demanding due to computational overheads during graph traversal in Dijkstra's Algorithm. In this study, three heap-based priority queues—Fibonacci Heap, D-ary Heap, and PriorityQueueHeap—are evaluated within Dijkstra's Algorithm for optimizing anomaly localization. Using GPerftools for monitoring memory allocation and for assessing execution time, this study determines the most suitable heap structure to utilize in managing large-scale dynamic graphs. The results show that heaps' selection greatly enhances the execution speed, memory utilization, and scalability in DGNN-based systems. Optimization enables real-time anomaly detection to be more effective, thus rendering critical infrastructure networks more resilient and secure.

Keywords

Graph Neural Network Real-Time Processing Dynamic Graph Heap Algorithm.

Publication Details

  • Type of Publication: Conference 
  • Conference Name: The 8th International Conference on Engineering Research, Innovation and Education (ICERIE 2025)
  • Date of Conference: 24/04/2025 - 24/04/2025
  • Venue: Shahjalal University of Science and Technology University Ave, Kumargaon, Sylhet 3114, Bangladesh
  • Organizer: School of Applied Sciences & Technology, Shahjalal University of Science and Technology
View on Publisher