Cloud Resource Utilization Prediction Using Runtime-Assembled Context-Specific Cooperative Artificial Neural Networks
Abstract
With the rapid expansion of cloud-based services in domains such as healthcare, finance, and AI, accurate resource utilization forecasting is essential for sustaining performance and reducing energy inefficiencies. We propose Runtime-Assembled Cooperative Artificial Neural Networks (RACANN), a modular, context-aware framework that dynamically assembles specialized models at runtime. By decomposing prediction tasks into lightweight ANN modules trained on specific temporal contexts (e.g., weekday, month), RACANN enhances prediction accuracy, reduces computational overhead, and improves interpretability. Experiments on real-world datasets show that while deep neural networks achieve slightly higher accuracy, RACANN offers superior transparency, efficiency, and modularity.
Department(s)
Computer Science
Document Type
Conference Proceeding
DOI
10.1109/ICCT67417.2025.11374113
Keywords
Artificial Neural Networks, Cloud, Machine Learning, Multi-Layer Perceptron, Resource Prediction
Publication Date
1-1-2025
Recommended Citation
Chowdhury, Shusmoy; Nelson, Nathan; Katangur, Ajay K.; Liu, Siming; and Saquer, Jamil M., "Cloud Resource Utilization Prediction Using Runtime-Assembled Context-Specific Cooperative Artificial Neural Networks" (2025). Faculty Scholarship. 206.
https://bearworks.missouristate.edu/articles00/206
Journal Title
International Conference on Communication Technology Proceedings ICCT
