Supermarket Sales Time Series Analysis

1. Project Overview
The owner of a mid-sized supermarket chain has been facing challenges in understanding and predicting his store's sales trends.
- Low sales during certain months despite consistent foot traffic.
- Overstocking some products while running out of others.
- Inability to predict sales during festive seasons, leading to missed opportunities.
To address these challenges, this project aims to analyze and forecast the daily sales revenue of the supermarket using time series forecasting techniques like ARIMA, SARIMA, and Prophet models.
2. Business Problem
- Understanding revenue patterns: Identifying why certain months show a drop in sales.
- Forecasting future sales: Planning for upcoming months to prevent over or under-stocking.
- Increasing profitability: Leveraging data-driven strategies to boost sales during low seasons.
3. Data Description
- Dataset: Supermarket sales data (one year)
- Key Features: InvoiceDate, Price
- Frequency: Daily
- Missing Values: Handled using forward fill method
- Outliers: Treated using the IQR method
4. Exploratory Data Analysis (EDA)
- Revenue distribution plots
- Monthly and quarterly trends
- Seasonal decomposition plots



5. Stationarity Tests
- ADF Test: Confirmed series is non-stationary.
- KPSS Test: Verified non-stationarity.
- Differencing applied to make the series stationary.
6. Forecasting Models
Explored different time series forecasting models to predict supermarket sales...
7. Model Comparison
Model | MAE | RMSE | MAPE (%) |
---|---|---|---|
ARIMA | 6464.07 | 10795.50 | 31.0% |
SARIMA | 4553.06 | 8944.56 | 23.0% |
Prophet | 3600 | 7500 | 18.0% |
8. Insights and Recommendations
- Peak Sales During Weekends: Enhanced staffing and inventory management needed.
- Higher Sales During Festive Seasons: Targeted marketing campaigns suggested.
- Product Preferences: Focus on popular items during holidays.
9. Conclusion
Successfully forecasted sales. Future work includes deep learning models (in progress).
10. Additional Resources
To dive deeper into time series forecasting techniques and gain a better understanding of the methods used in this project, you can explore the following resources:
Take a look at my other projects to see how I address real-world problems using data.