XGBoostPredict for Classification - Analytics Database

This example shows how to use XGBoost for classification: first use TD_XGBoost to train a classification model, and then use TD_XGBoostPredict to test the model.

The following sections show the input table, SQL query, and output model table when training a classification model using TD_XGBoost.

Input Table: xgboostTrain

SQL Function Call: TD_XGBoost

create table xgboost_classification_model AS(
    SELECT * FROM TD_XGBoost (
        ON xgboostTrain AS inputtable partition by ANY
        USING
        ResponseColumn('homestyle')
        InputColumns('rent','lotsize','bedrooms','bathrms','driveway')
        MaxDepth(3)
        NumBoostedTrees(1)
        ModelType('classification')
        Seed(1)
        IterNum(2)
    ) AS dt
)with data;

Output Model Table: xgboost_classification_model

The following sections show the input table, SQL query and output table when testing the classification model using TD_XGBoostPredict.

Input Table: xgboostTest

SQL Function Call: TD_XGBoostPredict Query

SELECT * FROM TD_XGBoostPredict(
    ON test_data AS inputtable partition by ANY
    ON xgboost_classification_model AS modeltable dimension
    USING
      IdColumn('id')
      ModelType('Classification')
      OutputProb('t')
      Responses('1','2','3')
) AS dt;

Output Table