This example calculates the cross-validation error for models trained by the GLML1L2 function by regression (the GLML1L2 call specified Family ('GAUSSIAN')) and shows the training error for each fold in the output table.
Input
The input table, admissions_train, is from GLM Example: Logistic Regression Analysis with Intercept.
SQL Call
SELECT * FROM CrossValidation2 ( ON admissions_train AS InputTable OUT TABLE OutputTable (cv_out) USING FunctionName ('GLML1L2') Metric ('mse') EvaluateTraining ('t') IdColumn ('id') TargetColumns ('masters', 'gpa', 'stats', 'programming') CategoricalColumns ('masters', 'stats', 'programming') ResponseColumn ('admitted') Family ('Gaussian') Alpha (0) RegularizationLambda (0.02) ) AS dt;
Output
fold_num | validation_mse | training_mse ---------------+-------------------+------------------- best_score | 0.225835152720764 | 0.15986538159187 average_score | 0.265838722846472 | 0.153415214510181 1 | 0.270134841041979 | 0.149599142261067 2 | 0.278042443533549 | 0.144148161584926 3 | 0.254015424730324 | 0.163766866292805 4 | 0.225835152720764 | 0.15986538159187 5 | 0.301165752205742 | 0.149696520820234
SELECT * FROM cv_out;
attribute | category | estimate | information ----------------+----------+--------------------+------------- (Intercept) | | 0.614277761352143 | p masters | yes | -0.377583967343473 | p stats | beginner | 0.198291660224301 | p stats | novice | 0.132269305463436 | p programming | beginner | -0.443020944585588 | p programming | novice | -0.204447536127055 | p gpa | | 0.0873461023294575 | p Family | | | Gaussian Regularization | | | Ridge Alpha | | 0 | Lambda | | 0.02 | Iterations # | | 16 | Converged | | | true Rows # | | 32 | Features # | | 7 | RMSE | | 0.409976691591614 | AIC | | 17.5666198827054 | BIC | | 27.8267712023035 |
Download a zip file of all examples and a SQL script file that creates their input tables.