Feature Engineering Techniques for Various Business Scenarios with Python Code Examples

Feature engineering is a crucial step in data analysis that transforms raw data into informative features to improve model performance.

1. E‑commerce scenario: Feature: order date; Technique: extract year, quarter, month, and weekday.

import pandas as pd
# Read order data
orders = pd.read_csv('orders.csv')
# Extract year
orders['Year'] = pd.to_datetime(orders['OrderDate']).dt.year
# Extract quarter
orders['Quarter'] = pd.to_datetime(orders['OrderDate']).dt.quarter
# Extract month
orders['Month'] = pd.to_datetime(orders['OrderDate']).dt.month
# Extract weekday
orders['Weekday'] = pd.to_datetime(orders['OrderDate']).dt.weekday

2. Market research scenario: Feature: text data; Technique: convert text to numeric vectors using bag‑of‑words or TF‑IDF.

from sklearn.feature_extraction.text import CountVectorizer
import pandas as pd
# Read survey data
survey_data = pd.read_csv('survey_data.csv')
# Extract text feature
text_data = survey_data['Response']
vectorizer = CountVectorizer()
text_features = vectorizer.fit_transform(text_data)

3. Healthcare scenario: Feature: patient age; Technique: calculate age from birth date.

import pandas as pd
import datetime
# Read patient data
patient_data = pd.read_csv('patient_data.csv')
# Calculate age
current_year = datetime.datetime.now().year
patient_data['Age'] = current_year - pd.to_datetime(patient_data['BirthDate']).dt.year

4. Finance scenario: Feature: time‑series data; Technique: create lag features such as previous close price and 7‑day average volume.

import pandas as pd
# Read stock data
stock_data = pd.read_csv('stock_data.csv')
# Extract lag feature
stock_data['PreviousClose'] = stock_data['Close'].shift(1)
stock_data['AverageVolume'] = stock_data['Volume'].rolling(window=7).mean()

5. Social media scenario: Feature: user registration date; Technique: compute usage duration as days since registration.

import pandas as pd
# Read user data
user_data = pd.read_csv('user_data.csv')
# Convert registration date
user_data['RegistrationDate'] = pd.to_datetime(user_data['RegistrationDate'])
# Calculate usage duration
user_data['UsageDuration'] = (pd.Timestamp.now() - user_data['RegistrationDate']).dt.days

6. Human resources scenario: Feature: employee hire date; Technique: extract month and quarter of hiring.

import pandas as pd
# Read employee data
employee_data = pd.read_csv('employee_data.csv')
# Extract month and quarter
employee_data['Month'] = pd.to_datetime(employee_data['HireDate']).dt.month
employee_data['Quarter'] = pd.to_datetime(employee_data['HireDate']).dt.quarter

7. Education scenario: Feature: student exam scores; Technique: compute average score and standard deviation per student.

import pandas as pd
# Read exam scores
exam_scores = pd.read_csv('exam_scores.csv')
# Calculate average and std deviation
exam_scores['AverageScore'] = exam_scores.mean(axis=1)
exam_scores['ScoreStd'] = exam_scores.std(axis=1)

8. Hotel scenario: Feature: booking and check‑in dates; Technique: calculate days in advance of booking.

import pandas as pd
# Read booking data
booking_data = pd.read_csv('booking_data.csv')
# Convert dates
booking_data['BookingDate'] = pd.to_datetime(booking_data['BookingDate'])
booking_data['CheckInDate'] = pd.to_datetime(booking_data['CheckInDate'])
# Calculate days in advance
booking_data['DaysInAdvance'] = (booking_data['CheckInDate'] - booking_data['BookingDate']).dt.days

9. Marketing scenario: Feature: ad clicks and impressions; Technique: compute click‑through rate.

import pandas as pd
# Read ad data
ad_data = pd.read_csv('ad_data.csv')
# Calculate CTR
ad_data['ClickThroughRate'] = ad_data['Clicks'] / ad_data['Impressions']

10. Logistics scenario: Feature: weight and volume of shipments; Technique: calculate density as weight divided by volume.

import pandas as pd
# Read shipment data
shipment_data = pd.read_csv('shipment_data.csv')
# Calculate density
shipment_data['Density'] = shipment_data['Weight'] / shipment_data['Volume']

The goal of feature engineering is to derive useful attributes from raw data to build more accurate and effective predictive models; the appropriate technique depends on the specific business context and data characteristics.