An artificial neural network is a powerful classification tool capable of generating complex boundaries between different classes of data.
The best choice of the number of neighbors depends upon the data; generally, larger values of neighbors reduce the effect of noise on the classification, but make boundaries between classes less distinct [ 15].
Neighbourhood size is essential in building the kNN classifier because it can strongly influence the quality of predictions; larger values of k can reduce the effect of noise on the classification but make boundaries between classes less distinct.
In any classification scheme, boundaries between classes are set arbitrarily; however, if the boundaries can be placed closest to natural divisions or gaps between classes, they will seem less random and subjective, and the standards will facilitate universal understanding.
Accuracy assessment is particularly problematic when thematic classes lie along a land-cover continuum, and boundaries between classes are ambiguous.
As such, individuals diagnosed under a certain diagnostic class are not clinically homogeneous, there are no clear boundaries between classes, and different classes are not mutually exclusive.
Decide between classes of rail travel.
The discrimination map should not, however, be interpreted as a representation of the statistical differences between classes but as a representation of the boundary between classes.
While this classifier does not apparently perform as well on other smaller and more diverse data sets, inconsistencies are largely related to intermediate cases and depend on where the boundary between classes is placed in a spectrum of cases in the training data.
Briefly, the SVM classifier principle is to find the optimal decision boundary between classes by maximizing the margin hyperplanes (the geometrical representation of the decision boundaries in multi-dimension) between the support vectors.
