Журнал «Современная Наука»

Russian (CIS)English (United Kingdom)
 MOSCOW +7(495)-142-86-81

HYBRID METHODS FOR PREDICTING THE EVOLUTION OF THE COMPONENT BASE OF COMPUTER SYSTEMS BASED ON THE INTEGRATION OF STATISTICAL AND NEURAL NETWORK MODELS

Maximenko Natalia Sergeevna  (Senior Lecturer, FSBEI HE "Donetsk National Technical University" )

Transition to heterogeneous nodes, chiplet integration, and coherent buses (CXL, PCIe 6.0) increases the cost of predictive errors in Perf/W, bandwidth, and reliability, while linear methods (ARIMA) are limited in nonlinear dynamics. The goal is to create a hybrid forecasting methodology for 1–5 year horizons through integration of SARIMA+LSTM+Transformer+GNN with Bayesian calibration and explainability (SHAP). Methods include robust-scaling, exogenous features (process technology, interfaces), quantile modeling, and rolling-origin validation. Results.The hybrid ensemble reduces RMSE by 15–24% and sMAPE by 20–30% relative to baseline SARIMA; intervals are calibrated (PICP=0.90–0.95, PINAW=0.15–0.20). Superiority is statistically significant (DM, p<0.05).

Keywords:forecasting, hybrid models, ARIMA/SARIMA, LSTM, Transformer for Time Series, graph neural networks, explainable AI, system analysis, Perf/W, SLA
 

Read the full article …

Citation link:
Maximenko N. S. HYBRID METHODS FOR PREDICTING THE EVOLUTION OF THE COMPONENT BASE OF COMPUTER SYSTEMS BASED ON THE INTEGRATION OF STATISTICAL AND NEURAL NETWORK MODELS // Современная наука: актуальные проблемы теории и практики. Серия: Естественные и Технические Науки. -2026. -№01. -С. 119-126 DOI 10.37882/2223-2966.2026.01.22
LEGAL INFORMATION:
Reproduction of materials is permitted only for non-commercial purposes with reference to the original publication. Protected by the laws of the Russian Federation. Any violations of the law are prosecuted.
© ООО "Научные технологии"

Terms Publications

Issues of the journals (Archive)

ECONOMICS AND LAW
HUMANITIES
NATURAL AND TECHNICAL SCIENCES
COGNITION

Серия - Natural and Technical Sciences