Recently, researchers have become increasingly interested in studying the dollar exchange rate because it impacts the economy and administration of countries. Real-time series data usually include linear and nonlinear components due to the presence of phenomena that are subject to mixed models. Based on these considerations, this study presents a hybrid model consisting of a linear model (MISO ARX) and a nonlinear model (GARCH-X). Thus, the hybrid model (MISO ARX-GARCH-X) was used to forecast the dollar exchange rates, as three methods were used to estimate the parameters of the linear model: (RLS-KF), (RELS-DC), and (RELS-TC), and a comparison was made between those methods based on comparison (MAE) and (MAPE). The (QMLE) method was used to estimate the parameters of the non-linear model (GARCH-X); one of the most important results reached is the superiority of the (RELS-DC) method over the rest of the methods in estimating the parameters of the (MISO ARX) model. The hybrid model (MISO ARX(1, 5, 3, 5, 3)-GARCH(2, 1)-X(0, 1)) also showed high accuracy in forecasting. The (MATLAB) program and the (R) program were used to analyze the data.