This article discusses the possibility of predicting the values of a series using complex-valued autoregression with an error for short-term forecasting. The authors consider the basic concepts of the function of a complex-valued variable and the model of complexvalued autoregression, together with the results of applying first- and second-order models of complex-valued autoregression with the CARE(p) error to describe and predict the initial series. The results obtained are compared with the first- and second-order autoregression in real numbers. A complex-valued autoregression model with an error showed a more accurate result for short-term forecasting, unlike the autoregression model in real numbers. The authors also conclude that complex-valued autoregression with an error is subject to further investigation in order to find out the prospects of using its imaginary part.