We used the IRIS dataset for this recipe, but we prepared the data ahead of time and then selected the distinct number of rows by using the NaiveBayesDataSet.distinct() API. We then proceeded to train the model using the NaiveBayes.train() API. In the last step, we predicted using .predict() and then evaluated the model performance via MulticlassMetrics() by outputting the confusion matrix, precision, and F-Measure metrics.
The idea here was to classify the observations based on a selected feature set (that is, feature engineering) into classes that correspond to the left-hand label. The difference here was that we are applying joint probability given conditional probability to the classification. This concept is known as ...
