Forecasting of a non-seasonal tourism time series with ANN

The paper present and discusses several alternative architectures of Artificial Neural Network models used to predict the time series of tourism demand for Cape Verde. This time series is particularly difficult to predict due to its non-seasonal characteristic usual in a similar time series for European Tourism destinations. The time index used as input and other input parameters variations improved the performance of the prediction over the test set to a relative error of 7.3% and a Pearson correlation coefficient of 0.92.