After remaining a long time in academic circles due to their advanced mathematical background, neural networks (NN) rapidly grew in popularity as more practically usable formats are available – like the built-in function of R. NNs are artificial intelligence adaptive software that can detect complex patterns in data: it is just like an old trader who has a good market intuition but cannot always explain to you why he is convinced you should go short on the Dow Jones Industrial Average index (DIJA).

The network architecture consists of a number of nodes connected by links. Networks usually have 3 or 4 layers: input, hidden and output layers, and in each layer several neurons can be found. The number of first layer's nodes corresponds to the number of the model's explanatory variables, while the last layer's equals to the number of the response variables (usually 2 neurons in case of binary target variable or 1 neuron in case of continuous target variable). The model's complexity...