Execute lightGBM functions | teradataml open-source machine learning functions - Execute lightGBM functions using standard supported arguments - Teradata Package for Python

Using training APIs for training

To run training APIs train and cv, create a Dataset object then pass the object as input to these functions. The following subsections cover both single model and multi model training and highlight the differences between training with actual lightGBM and td_lightgbm.

Whether it is single-model training or multi-model training, you must first create the required teradataml DataFrames. These DataFrames should be created from the same parent teradataml DataFrame using the select() API.

# Load the example dataset.
>>> load_example_data("openml", ["multi_model_classification", "multi_model_regression"])

## Classification data.
# Create DataFrames.
>>> df_train_classif = DataFrame('multi_model_classification')
>>> df_train_classif
        col1	    col2	      col3	        col4  label	group_column partition_column_1	partition_column_2
-2.86976401	 -1.45650077  -0.096175758	-1.831834743 	  0	          11	              1	                11
-2.69651383	-1.244829178  -0.110505514	-1.670524168	  0	           9	              0	                10
0.939206169	-0.618203487   0.209642108	 0.150728526	  0	          11	              0	                10
1.413963995	 0.329342266   0.110572078	 0.743405742	  0	          10	              0	                11
0.599510805	 1.298906549  -0.141761534	 0.790378168	  1	           8	              1	                10
-1.17273888	  0.05006965  -0.144305532	-0.484090357	  1	          12	              1	                11
1.087219107	-1.006162386   0.289957882	 0.055393647	  0	           9	              1	                10
0.669401117	 0.965276402  -0.079356706	 0.683697322	  1	           8	              1	                10
1.441799942	-0.007863088    0.16867607	 0.617164007	  0	           9	              1	                10
1.619580063	 0.479070866   0.110079835	 0.893252734	  1	           8	              0	                10

# Create required classification DataFrames.
>>> df_x_classif = df_train_classif.select(["col1", "col2", "col3", "col4"])
>>> df_y_classif = df_train_classif.select("label")

## Regression data.
# Create DataFrames.
>>> df_train_reg = DataFrame('multi_model_regression')
>>> df_train_reg.head(4)
        col1	    col2	      col3	        col4  label	group_column partition_column_1	partition_column_2
-2.86976401	 -1.45650077  -0.096175758	-1.831834743   -141	          11	              1	                11
-2.69651383	-1.244829178  -0.110505514	-1.670524168    -89            9	              0	                10
0.939206169	-0.618203487   0.209642108	 0.150728526   -137	          11	              0	                10
1.413963995	 0.329342266   0.110572078	 0.743405742	  8	          10	              0	                11

# Create required regression DataFrames.
df_x_reg = df_train_reg.select(["col1", "col2", "col3", "col4"])
df_y_reg = df_train_reg.select("label")