Enhancing COVID-19 Forecasting in Dagestan with Quasi-linear Recurrence Equations by using GLDM Algorithm
Mostafa Salaheldin Abdelsalam Abotaleb, Tatiana Makarovskikh, Maad Mohsin Mijwil, Ramadhan Ali J · Al-Bahir Journal for Engineering and Pure Sciences · 2024
This research delineates the advancement of a refined predictive algorithm centered on the Generalized Least Deviation Method (GLDM) specifically configured for analyzing COVID-19 infection trends in Dagestan using univariate time series data. Our methodology is characterized by its enhancement of forecast precision through diligent minimization of a bespoke loss function. The algorithm’s innovation lies in its formulation, incorporating second-order relationships within the time series data: where are the computed weights ascribed to historical data, and denotes the error component. Our empirical analysis substantiates that by strategically accentuating pertinent coefficients and optimizing the loss function, there is a significant elevation in the model’s forecasting accuracy. Consequently, the refined second-order GLDM model emerges as an advanced and applicable instrument for the prognostication of COVID-19 infection cases in Dagestan.