r-AutoNN

Provides optimal combinations of input nodes and hidden neurons for fitting feedforward single-layer artificial neural networks in time series forecasting. Models are evaluated using root mean square error, mean absolute percentage error, and mean absolute error measures.