Skip to content

EDA Workflow

🎯 Purpose

This checklist provides a high-level workflow for conducting exploratory data analysis (EDA). It covers the essential steps from initial data inspection to pre-modeling cleanup.

Related Guides

This checklist is a companion to the 📊 General EDA Guidebook, 📊 Visual EDA Interpretation Guide, and 📘 Advanced EDA Guidebook.

📦 1. Dataset Structure

  • [ ] .shape, .info() — Dimensions & column types
  • [ ] .describe(include='all') — Descriptive stats
  • [ ] .duplicated().sum() — Check for row duplication
  • [ ] .memory_usage() — Memory optimization

🔍 2. Variable Exploration

Numeric

  • [ ] Histograms + KDE
  • [ ] Skew, kurtosis
  • [ ] Outlier detection (boxplot, z-score, IQR)

Categorical

  • [ ] Frequency counts
  • [ ] Unique counts / high cardinality flags
  • [ ] Encoding strategy (if needed)

📈 3. Distribution & Relationship Plots

  • [ ] Boxplot by category
  • [ ] Correlation heatmap (numeric only)
  • [ ] Pairplot or scatter matrix
  • [ ] Crosstabs (for categorical pairs)

🧪 4. Missing Values

  • [ ] df.isnull().sum() summary
  • [ ] Heatmap of missingness
  • [ ] Consider: drop, fill, predictive imputation

📉 5. Outliers

  • [ ] IQR filtering
  • [ ] Z-score filtering
  • [ ] Context/domain filtering (caps/floors)

🔁 6. Scaling & Transformation

  • [ ] Log/sqrt transforms for skew
  • [ ] StandardScaler, MinMaxScaler, RobustScaler

🧠 7. Feature Engineering (Optional)

  • [ ] Bin continuous vars (if needed)
  • [ ] Polynomial or interaction terms
  • [ ] Flag rare categories or consolidate groups

🧹 8. Pre-Model Cleanup

  • [ ] Drop ID / index / time features if not useful
  • [ ] Recheck all dtypes
  • [ ] Align feature set with model goals

🧠 Final Tip

"EDA is an iterative process. Don't be afraid to circle back to earlier steps as you uncover new insights."

Ready to export to modeling script or notebook? ✅