NaiveBayesTextClassifierTrainer Example - Teradata Vantage

Input

• tokens: Created by applying the TextTokenizer function to the training table complaints, a log of vehicle complaints

  In complaints, the category column indicates whether the car has been in a crash.

SQL Call

This call creates the model table, complaints_tokens_model, by calling NaiveBayesTextClassifierTrainer. It creates the NaiveBayesTextClassifierTrainer input table, token, by applying TextTokenizer to the table complaints.

CREATE MULTISET TABLE complaints_tokens_model AS (
  SELECT * FROM NaiveBayesTextClassifierTrainer (
    ON (
      SELECT * FROM NaiveBayesTextClassifierInternal (
        ON (
          SELECT doc_id, lower(token) AS token, category
          FROM TextTokenizer (
            ON complaints PARTITION BY ANY
            USING
            TextColumn ('text_data')
            OutputByWord ('true')
            Accumulate ('doc_id', 'category')
          ) AS dt1
        ) AS "input" PARTITION BY category
        USING
        TokenColumn ('token')
        ModelType ('Bernoulli')
        DocIDColumns ('doc_id')
        DocCategoryColumn ('category')
      ) AS dt2
    ) PARTITION BY 1
  ) AS dt3
) WITH DATA;

Output

This query returns the following table:

SELECT * FROM complaints_tokens_model;