Analysis and projection of Pfizer's stock returns, in the period 2018-2020, through differential neural networks

Abstract

In this paper, a differential neural network (DNN) is used to project Pfizer’s stock returns in the 2018-2020 period. The model uses quarterly data, at the end of the period, the price of the company’s stock (P), net sales (NS), total assets (TA) and accounts receivable (AR). The results are compared with the classic regression models and there is evidence of the superior goodness of fit of the DNN, compared to conventional methods, since the error in out sample forecast is less than 5 %.