On Exponential Vector Autoregression Method for Data Transmission Channel Quality Forecasting

This work presents a modified vector autoregression (VAR) method for forecasting overlay channel quality metrics. The modification introduces weight coefficients for time series quantiles, with an exponential weighting approach proposed to calculate these coefficients. Experimental results demonstrate that this modified approach achieves a prediction accuracy improvement of 2.49\% to 12.45\% compared to classical AR and VAR methods.