M5 Forecasting: Walmart Sales Exploratory Data Analysis
Data Science Time Series Kaggle EDA
A comprehensive exploratory analysis of the M5 Forecasting competition dataset, analyzing Walmart’s daily sales across multiple product categories, departments, stores, and states.
Key Insights
- Intermittent Demand: Many products exhibit zero-sales days, requiring specialized intermittent demand forecasting models like Croston’s method.
- Seasonality & Events: Sales patterns show strong weekly seasonality, annual cycles, and sharp spikes/dips during calendar events like Thanksgiving and Christmas.
- Price Elasticity: Promotion flags and price changes have a direct, measurable impact on sales velocity across different categories.
Check out the complete code, visualizations, and modeling setup in the Kaggle notebook.
Kaggle Notebook
Explore the Full Analysis
Access the complete Kaggle notebook to view all source code, interactive visualizations, and model configurations.