You work for a retail clothing manufacturer that has a very active online web store. You have been assigned the task of building a model to contact customers for a direct marketing campaign based on their predicted receptiveness to the campaign. Some of your customers have been contacted in the past for other marketing campaigns. You don’t want to contact these customers who have been contacted in the past for this latest campaign.
Before training this model, you need to clean your data and prepare it for the XGBoost algorithm you are going to use. You have written your cleaning/preparation code in your SageMaker notebook. Based on the following code, what happens on lines 19, 21, 22? (Select THREE)
1 import sagemaker
2 import boto3
3 from sagemaker.predictor import csv_serializer 
4 import numpy as np
5 import pandas as pd
6 from time import gmtime, strftime
7 import os
 8 region = boto3.Session().region_name 
9 smclient = boto3.Session().client(`sagemaker`)
10 from sagemaker import get_execution_role
11 role = get_execution_role()
12 bucket = `sagemakerS3Bucket` 
13 prefix = `sagemaker/xgboost`
14 !wget -N https://.../bank.zip
15 !unzip -o bank.zip
16 data = pd.read_csv(`./bank/bank-full.csv`, sep=`;`)
17 pd.set_option(`display.max_columns`, 500) 
18 pd.set_option(`display.max_rows`, 5) 
19 data[`no_previous_campaign`] = np.where(data[`contacted`] == 999, 1, 0)
20 data[`not_employed`] = np.where(np.in1d(data[`job`], [`student`, `retired`, `unempl`]), 1, 0)
21 model_data = pd.get_dummies(data) 
22 model_data = model_data.drop([`duration`, `employee.rate`, `construction.price.idex`,
`construction.confidence.idx`,`lifetime.rate`, `region`], axis=1)
23 train_data, validation_data, test_data = np.split(model_data.sample(frac=1,
random_state=1729), [int(0.7 * len(model_data)), int(0.9*len(model_data))]) 
24 pd.concat([train_data[`y_yes`], train_data.drop([`y_no`, `y_yes`], axis=1)],
axis=1).to_csv(`train.csv`, index=False, header=False)
25 pd.concat([validation_data[`y_yes`], validation_data.drop([`y_no`, `y_yes`], axis=1)],
axis=1).to_csv(`validation.csv`, index=False, header=False)
26 pd.concat([test_data[`y_yes`], test_data.drop([`y_no`, `y_yes`], axis=1)],
axis=1).to_csv(`test.csv`, index=False, header=False)
27 boto3.Session().resource(`s3`).Bucket(bucket).Object(os.path.join(prefix,
`train/train.csv`)).upload_file(`train.csv`)
28 boto3.Session().resource(`s3`).Bucket(bucket).Object(os.path.join(prefix,
`validation/validation.csv`)).upload_file(`validation.csv`)

Answer options:

A.Splits bank dataset into train, validation, and test datasets
B.Sets the attribute no_previous_campaign to 999, 0, or 1 depending if the customer in the observation has been contacted via a previous campaign
C.Sets the attribute no_previous_campaign to 1 if the customer in the observation has not been contacted via a previous campaign or 0 if they have been contacted via a previous campaign
D.Converts categorical data to a set of indicator variables
E.Converts empty attributes to dummy variables
F.Removes features deemed inconsequential
G.Removes observations deemed inconsequential

Answers: C, D, F
Option A is incorrect. This option describes what happens on line 23, not what happens on lines 20, 21, or 22.
Option B is incorrect. Line 19 does not set the attribute no_previous_campaign to 999. It sets the attribute no_previous_campaign to 1 or 0 depending on whether the customer in the observation has been contacted via a previous campaign, as indicated by the value 999.
Option C is correct. Line 19 sets the attribute no_previous_campaign to 1 or 0 provided the customer in the observation has been contacted via a previous campaign, as indicated by the value 999.
Option D is correct. Line 21 uses the pandas library get_dummies method to convert the categorical attributes in the dataframe to dummy (or indicator) variables.
Option E is incorrect. Line 21 does not convert empty attributes to dummy variables. It uses the pandas library get_dummies method to convert the categorical attributes in the dataframe to dummy (or indicator) variables.
Option F is correct. Line 22 removes (or drops) several features, presumably because you have deemed the features inconsequential to the training of your model.
Option G is incorrect. Line 22, in this usage, calls the pandas drop method to remove features, not observations.
Reference:
Please see the SciPy numpy.where documentation (for line 19), the pandas get_dummies documentation (for line 21), and the pandas DataFrame.drop documentation (for line 22)