Table 3. Naïve Bayes classifier.
Prior Probabilities | Uniform | Empirical | ||
Distribution | Training set | Test set | Training set | Test set |
Normal | 47.47 | 58.00 | 46.78 | 57.28 |
Kernel | 33.45 | 56.98 | 33.04 | 57.36 |
Training set and test set errors are shown for each combination of type of distribution of data and prior class probabilities. Lowest error rate is ∼33% for training set and ∼57% for test set when kernel distribution is used to model the data.