Skip to content

Lab 01: SageMaker Data Wrangler

Domain: 1 - Data Preparation
Difficulty: Easy
Time: 30 minutes

Objective

Learn to use SageMaker Data Wrangler for visual data preparation and feature engineering.

Prerequisites

  • SageMaker Studio access
  • Sample dataset (CSV)

Steps

Step 1: Access Data Wrangler

  1. Open SageMaker Studio
  2. From the launcher, select "New data flow"
  3. Name your flow: customer-churn-prep

Step 2: Import Data

  1. Click "Import data"
  2. Select "Amazon S3"
  3. Navigate to your dataset or use a sample: 
    s3://sagemaker-sample-files/datasets/tabular/synthetic/churn.csv
  4. Click "Import"

Step 3: Explore Data

  1. Click on the dataset node
  2. Select "Add analysis"
  3. Choose "Table summary" to view statistics
  4. Create a "Histogram" for numerical columns

Step 4: Add Transformations

  1. Click "+" after the data node
  2. Select "Add transform"
  3. Apply these transformations:
  4. Handle missing: Fill missing values
  5. Encode categorical: One-hot encode State
  6. Drop columns: Remove Phone
  7. Custom transform: Create new feature
# Custom Pandas transform
df['HighUsage'] = (df['Day Mins'] > df['Day Mins'].median()).astype(int)

Step 5: Export Flow

  1. Click "Export"
  2. Choose export destination:
  3. "Export to S3" for data
  4. "Export to Pipeline" for automation
  5. "Export to Python code" for reuse

Verification

  • Transformed dataset exported to S3
  • Data quality improved (no missing values)
  • Features properly encoded

Cleanup

  1. Close Data Wrangler flow
  2. Delete exported files if not needed
  3. Stop Studio instance

Key Takeaways

!!! note "Exam Points" - Data Wrangler is visual, no-code data preparation - Supports 300+ built-in transformations - Can export to Pipelines for automation - Part of SageMaker Studio