Table 2.

Parametrization of algorithms.

Algorithm	Parameters
kNN	k = 1 and Euclidian distance
SOM	Euclidian distance, batch training, maximum training time equal to 1000, rectangular lattice, and Gaussian neighborhood function with maximum aperture of 1 with decay due to the number of iterations. The SOM map dimension has the square root of the number of dataset objects by two ($\sqrt{N}/\mathrm{2}$)
iNN	Execution of the kNN algorithm with k value equal to 7 (best result from [10]) and informative neighbor number equal to 1