Add a Naive Bayes classifier

[ChristophWurst]

... because literature sometimes suggest this for intrusion detection systems, so I figured I'd give that a shot as well.

Results are worse and more fluctuating than the ones from the MLP classifier, but it's impressively fast.

Got 4452 samples for training: 636 positive, 1908 random negative and 1908 shuffled negative
Got 113 positive and 113 negative samples for validation (rate: 0.15)
Vecor dimensions: 48
Start training
Training finished after 0s

Run predictions on test data set
Predictions calculated

Persisting trained model
Model 39 persisted
Prescision(y): 0.875
Prescision(n): 0.67532467532468
Recall(y): 0.55752212389381
Recall(n): 0.92035398230088

Note: this model assumes statistical independence of featues and I'm not sure that is given with the current representation of input values.