In this project, the predictive accuracy of statistical models and machine learning methods applied to five hourly time series of sales quantity from a retail e-commerce was investigated. The selected models included dynamic regression estimated by ordinary least squares (OLS), Lasso, and AdaLasso, in addition to the random forest method. Predictive accuracy was assessed for forecast horizons ranging from 1 to 12 hours ahead, using the MAE and RMSE metrics. The results indicated that models from the Lasso family exhibited superior performance according to the MAE metric. Regarding RMSE, the best results were associated with the dynamic regression model that incorporates autoregressive terms of the sales quantity and dummy variables (RegrDin(3)). The computational implementation of the models was carried out using the programming languages Python and R.