Naive Bayes Classifier Example - Teradata Vantage

Input

The input table of raw data, nb_input_iris, contains values for four attributes (sepal_length, sepal_width, petal_length and petal_width), grouped into three categories (setosa, versicolor and virginica).

This code divides the 150 data rows into a training data set (80%) and a testing data set (20%):

DROP TABLE nb_iris_input_train;
DROP TABLE nb_iris_input_test;

CREATE MULTISET TABLE nb_iris_input_train AS (
  SELECT * FROM nb_input_iris WHERE id MOD 5 <> 0
) WITH DATA;

CREATE MULTISET TABLE nb_iris_input_test AS (
  SELECT * FROM nb_input_iris WHERE id MOD 5 = 0
) WITH DATA;

This query returns the following table:

SELECT * FROM nb_iris_input_train ORDER BY id;

This query returns the following table:

SELECT * FROM nb_iris_input_test ORDER BY id;

SQL Call

DROP TABLE nb_iris_model;

CREATE MULTISET TABLE nb_iris_model AS (
  SELECT * FROM NaiveBayesReduce (
    ON NaiveBayesMap (
      ON nb_iris_input_train
      USING
      ResponseColumn ('species')
      NumericInputs ('[1:4]')
    ) PARTITION BY class_nb
  ) AS dt 
) WITH DATA;

Output

This query returns the following table:

SELECT * FROM nb_iris_model ORDER BY 1;