Pressure drop time series from a fluidized bed are used for the evaluation of chaotic invariants, such as the Kolmogorov entropy and Lyapunov exponents. They are sensitive to noisy data and require a minimum number of points to obtain them. Therefore, a good forecast of chaotic time series could provide an estimative for these parameters. This work compares the prediction performance of two different methods, the popular ARIMA model and the k-nearest neighbor algorithm. The datum set used is an experimental minimum fluidization pressure drop time series, in which the fluid phase was atmospheric air and the particle phase was glass beads belonging to the B group of the Geldart classification. The comparison was made evaluating the root mean squared error of each forecast. It was verified that the k-nearest neighbor algorithm captures well the dynamics of the series while the ARIMA models does not. The two methods root mean squared error differ by three orders of magnitude. Therefore, relevant information can be extracted from the forecast of a pressure drop time series using the k-nearest neighbor algorithm, as it gives a reliable prediction.