2.2.2. Variable selection
Since the inclusion of all the explanatory variables may decrease the accuracy of the model by introducing noise, the filter-wrapper approach was followed [5]. It is based on mutual information between the variables, which are included one by one in the model provided that the new variable improves the accuracy of the model; otherwise, it is deleted. The accuracy of the model is measured by the root mean squared error, which represents the difference between observed and predicted values, obtained as