Score the Predictive Bear Marketing Model with APPLY Table Operator | Teradata Open Analytics Framework - Score the Prediction Model - Teradata VantageCloud Lake

#!/usr/bin/env python3
import sys
import pandas as pd
import pickle
import xgboost as xgb
def load_data():
    try:
        return pd.read_csv(sys.stdin, delimiter=",", header=None)
    except Exception as e:
        print(e, file=sys.stderr)
        sys.exit(0)
def main():
    df = load_data()
    print(f"Number of columns in df: {len(df.columns)}\n", file=sys.stderr)
    print(f"Number of rows in df: {len(df)}\n", file=sys.stderr)
    if df.empty:
        sys.exit(0)
    category_columns_indices = [1,2,3,4,6,7,8,10,15]
    for col in category_columns_indices:
        df[col] = df[col].astype('category')
    test = df
    x_test = test.iloc[:, 0:15]
    y_test = test.iloc[:, 16]
    y_test = y_test.map({'yes': 1, 'no': 0})
    dtest_reg = xgb.DMatrix(x_test, y_test, enable_categorical=True)
    model_file_name = "xgb_model_"
    model_xgb = xgb.Booster()
    model_xgb.load_model(model_file_name)
    predict = model_xgb.predict(
        dtest_reg,
    )
    preds_df = pd.DataFrame(predict, columns=['Predictions'])
    x_test['Predictions'] = preds_df['Predictions'].values
    x_test['Predictions_str'] = x_test['Predictions'].apply(lambda x: 'Yes' if x >= 0.5 else 'No')
    csv_string = x_test.to_csv(index=False, header=False)
    # Remove the last line from the CSV string, it is empty.
    csv_string = '\n'.join(csv_string.split('\n')[:-1])
    print(csv_string, file=sys.stdout)
if __name__ == "__main__":
    main()

from teradataml import Apply
display.print_sqlmr_query = True
apply_obj = Apply(data=bank_df_test,
                 data_partition_column="partition_column_1",
                 apply_command='python3 bank-marketing-predict.py',
                 returns={"age": INTEGER(), "job": VARCHAR(), "marital": VARCHAR(), "education": VARCHAR(), "default_value": VARCHAR(), "balance": INTEGER(), "housing": VARCHAR(), "loan": VARCHAR(), "contact": VARCHAR(), "day_of_month": INTEGER(), "month_of_year": VARCHAR(), "duration": INTEGER(), "campaign": INTEGER(), "pdays": INTEGER(), "previous": INTEGER(), "prediction" : FLOAT(), "Predictions_str" : VARCHAR()}
                )
prediction = apply_obj.execute_script()

Out:

SELECT * FROM Apply(
	ON "ALICE"."ml__select__1714117364845315" AS "input"
	PARTITION BY "partition_column_1"
	returns(age INTEGER, job LONG VARCHAR, marital LONG VARCHAR, education LONG VARCHAR, default_value LONG VARCHAR, balance INTEGER, housing LONG VARCHAR, loan LONG VARCHAR, contact LONG VARCHAR, day_of_month INTEGER, month_of_year LONG VARCHAR, duration INTEGER, campaign INTEGER, pdays INTEGER, previous INTEGER, prediction FLOAT, Predictions_str LONG VARCHAR)
	USING
	APPLY_COMMAND('python3 bank-marketing-predict.py')
	ENVIRONMENT('bank-marketing-env')
	STYLE('csv')
	delimiter(',') 
) as sqlmr

prediction

Out:

age 	job 	marital 	education 	default_value 	balance 	housing 	loan 	contact 	day_of_month 	month_of_year 	duration 	campaign 	pdays 	previous 	prediction 	Predictions_str
30 	management 	married 	tertiary 	no 	1567 	yes 	no 	cellular 	12 	aug 	1133 	4 	-1 	0 	1.0 	Yes
30 	services 	divorced 	secondary 	no 	1336 	yes 	no 	cellular 	18 	may 	13 	3 	355 	5 	0.0 	No
30 	unemployed 	divorced 	secondary 	no 	7105 	no 	no 	cellular 	22 	dec 	321 	2 	-1 	0 	1.0 	Yes
30 	blue-collar 	married 	primary 	no 	5 	no 	no 	cellular 	23 	dec 	282 	1 	203 	1 	1.0 	Yes
30 	management 	married 	tertiary 	no 	102 	yes 	no 	cellular 	5 	apr 	470 	7 	426 	3 	1.0 	Yes
30 	technician 	single 	tertiary 	no 	3086 	yes 	no 	cellular 	7 	may 	84 	3 	168 	1 	0.0 	No
30 	technician 	married 	secondary 	no 	2278 	no 	yes 	telephone 	19 	nov 	47 	1 	-1 	0 	0.0 	No
30 	services 	married 	tertiary 	no 	81 	no 	no 	cellular 	22 	apr 	145 	1 	-1 	0 	1.0 	Yes
30 	technician 	single 	tertiary 	no 	411 	no 	no 	cellular 	22 	oct 	127 	3 	149 	2 	1.0 	Yes
30 	management 	single 	tertiary 	no 	168 	no 	no 	cellular 	27 	aug 	35 	13 	-1 	0 	0.0 	No