On the prediction of chaotic time series using neural networks

Prediction techniques have the challenge of guaranteeing large horizons for chaotic time series. For instance, this paper shows that majority can predict one step ahead with relatively low root-mean-square error (RMSE) and Symmetric Mean Absolute Percentage Error (SMAPE). However, some based on neural networks more steps similar RMSE SMAPE values. In manner, work provides a summary prediction techniques, including type series, predicted ahead, error. Among those echo state network (ESN), long short-term memory, artificial convolutional are compared conditions to up ten Lorenz-chaotic The comparison among these include values, training testing times, required memory in each case. Finally, considering SMAPE, few neurons reservoir, performance an ESN is good technique five fifteen using thirty taking lowest tracking cases.