Practical Data Sampling Techniques and Code Examples for Various Business Scenarios

Data sampling is a commonly used technique in data analysis that can improve efficiency while preserving representativeness of the original data.

1. E‑commerce scenario: Randomly sample a subset of user purchase records for analysis.

import pandas as pd
# Read purchase records
purchase_data = pd.read_csv('purchase_data.csv')
# Randomly sample 1,000 records
sample_data = purchase_data.sample(n=1000)

2. Market research scenario: Use stratified sampling based on respondent attributes such as gender.

# Read survey data
survey_data = pd.read_csv('survey_data.csv')
# Stratified sampling by gender
male_data = survey_data[survey_data['Gender'] == 'Male'].sample(n=500)
female_data = survey_data[survey_data['Gender'] == 'Female'].sample(n=500)

3. Healthcare scenario: Apply a time‑window sample to select records within a specific year.

# Read medical records
medical_data = pd.read_csv('medical_data.csv')
# Time‑window sampling for 2022
sample_data = medical_data[(medical_data['Date'] >= '2022-01-01') & (medical_data['Date'] <= '2022-12-31')]

4. Finance scenario: Use a sliding‑window sample to take the most recent 30 days of stock trading data.

# Read stock data
stock_data = pd.read_csv('stock_data.csv')
# Sliding‑window sample: last 30 days
sample_data = stock_data.tail(30)

5. Social media scenario: Keyword sampling selects comments containing a specific keyword.

# Read comment data
comment_data = pd.read_csv('comment_data.csv')
# Keyword sampling for "好评"
sample_data = comment_data[comment_data['Content'].str.contains('好评')]

6. Human resources scenario: Group sampling picks a proportion of employees from each department.

# Read performance data
performance_data = pd.read_csv('performance_data.csv')
# Group sampling: 10% from each department
sample_data = performance_data.groupby('Department').apply(lambda x: x.sample(frac=0.1))

7. Education scenario: Stratified random sampling selects a percentage of students from each grade.

# Read exam scores
exam_scores = pd.read_csv('exam_scores.csv')
# Stratified random sampling: 20% per grade
sample_data = exam_scores.groupby('Grade').apply(lambda x: x.sample(frac=0.2))

8. Hotel scenario: Interval sampling picks records at regular time intervals (e.g., every week).

# Read booking data
booking_data = pd.read_csv('booking_data.csv')
# Interval sampling: every 7th record (weekly)
sample_data = booking_data[::7]

9. Marketing scenario: Click‑based sampling selects ads with the highest number of clicks.

# Read ad data
ad_data = pd.read_csv('ad_data.csv')
# Click‑based sampling: top 100 ads by clicks
sample_data = ad_data.nlargest(100, 'Clicks')

10. Logistics scenario: Weight‑based sampling chooses shipments with the largest weight, using a quantile threshold.

# Read shipment data
shipment_data = pd.read_csv('shipment_data.csv')
# Determine 90th percentile weight threshold
threshold = shipment_data['Weight'].quantile(0.9)
# Weight‑based sampling: shipments >= threshold
sample_data = shipment_data[shipment_data['Weight'] >= threshold]

These examples demonstrate how different sampling strategies can reduce data volume, speed up analysis, and maintain the representativeness of the original dataset across diverse business domains.