2026-05-04T22:06:53.980Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:11517 2025-08-28T22:49:32Z 3

Forecasting Cloud Instance Pricing Using Metaheuristic Optimizer Multi-headed Recurrent Neural Networks, Chapter in LNNS Lecture Notes in Networks and Systems: ICDMIS 2024: Data Mining and Information Security, Springer, volume 1388 2025 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/11517 https://link.springer.com/chapter/10.1007/978-981-96-6063-6_31 M. Protic L. Jovanovic M. Mravik N. Bacanin J. Perisic M. Zivkovic M. Antonijevic Modern systems need dependable processing and hosting in order to function as businesses become more and more reliant on cloud infrastructure. However, because cloud service pricing structures are changeable and based on both particular company demands and broader market demand, it can be challenging for businesses to precisely forecast costs. In order to improve the accuracy of cloud instance pricing evaluations and projections, this research explores the usage of multi-headed recurrent architectures, offering a solid framework for more informed budgeting. Understanding that hyperparameter selection plays a major role in recurrent neural network (RNN) performance, this work employs a number of modern optimization techniques and presents an altered version of a popular optimizer to overcome certain constraints. The results of this optimized model record the mean absolute error (MAE) of 0.000830 and it can be said that the application itself is realistic for predicting the price of cloud services. bookPart Springer, Singapore 471 484 10.1007/978-981-96-6063-6_31 LNNS Lecture Notes in Networks and Systems: ICDMIS 2024: Proceedings of Data Mining and Information Security, volume 1388