Load Data for Training | Scoring with a micro model file | Open Analytics Framework - Load Data for Training - Teradata Vantage

fit_input_table  = DataFrame(table_name='sampled_fit_table')

import pandas as pd
import statsmodels.api as sm
import numpy as np
import sys
import pickle
import warnings
import random
import string

warnings.filterwarnings("ignore")

DELIMITER=','

# Know your data: You must know in advance the number and data types of the
# incoming columns from the SQL Engine database!
# For this script, the input expected format is:
# 0: p_id, 1-5: indep vars, 6: dep var, 7: nRow, 8: model (if nRow==1), NULL o/w
colNames = ['p_id','x1','x2','x3','x4','x5','y']

# All input columns are float numbers.
# If any numbers are streamed in scientific format that contains blanks i
# (such as "1 E002" for 100), the following Lambda function removes blanks
# from the input string so that Python interprets the number correctly.
sciStrToFloat = lambda x: float("".join(x.split()))
# Use the Lambda function in the following converters for each input column.
converters = {0: sciStrToFloat,
              1: sciStrToFloat,
              2: sciStrToFloat,
              3: sciStrToFloat,
              4: sciStrToFloat,
              5: sciStrToFloat,
              6: sciStrToFloat}

try:
    ### Ingest and process the rest of the input data rows
    ###
    input_file = sys.stdin
    df = pd.read_csv(input_file, sep=DELIMITER, header=None, names=colNames,
                     index_col=False, iterator=False, converters=converters)

    # Print the number of rows and columns in the dataframe
    #print(f"Number of rows in df: {len(df)}", file=sys.stderr)
    #print(f"Number of columns in df: {len(df.columns)}", file=sys.stderr)

    # For AMPs that receive no data, exit the script instance gracefully.
    if df.empty:
        sys.exit(0)

    # Create object with intercept and independent variables. The intercept column
    # must be present to use the object in the StatsModels GLM() in the following.
    dfx = df.loc[:,'x1':'x5']
    dfx.insert(0,'Intercept',1.0)
    # Create object with dependent variable
    dfy = df.loc[:,'y']
    # Use GLM in statsmodels for binomial general linear modeling.
    logit = sm.GLM(dfy, dfx, family = sm.families.Binomial())

    # Fit the model. Use disp=0 in the parenthesis to prevent sterr output.
    fitResult = logit.fit(disp=0)

    model_name = ''.join(random.choices(string.ascii_letters + string.digits, k=32))
    model_path = f"/lob/{model_name}.pkl"

    with open(model_path, "wb") as model_file:
        pickle.dump(fitResult, model_file)

    # Export results to the SQL Engine database through standard output
    print('{}{}{}'.format(df.loc[0]['p_id'], DELIMITER, model_path))
except Exception as e:
    print(f"Error : {e}")
    sys.exit(1)