ML-powered sales forecasting built with Python on the Superstore dataset.
👉 https://nikhil-ml-sales-forecast.netlify.app/
Predicts monthly sales using 3 ML models trained on 4 years of Superstore retail data (2014–2017) with 13 engineered temporal features. Includes a fully interactive web dashboard.
| Model | MAE | RMSE | R² | MAPE |
|---|---|---|---|---|
| Linear Regression ⭐ | $12,293 | $15,092 | 0.600 | 16.8% |
| Random Forest | $14,237 | $16,902 | 0.493 | 19.7% |
| Gradient Boosting | $15,586 | $16,591 | 0.511 | 22.9% |
✅ Best Model: Linear Regression (lowest MAPE 16.8%)
| File | Description |
|---|---|
index.html |
Interactive web dashboard |
sales_forecasting_complete.ipynb |
Full ML pipeline notebook |
Sample - Superstore.csv |
Raw dataset (9,994 rows) |
Python · Pandas · NumPy · Scikit-learn · Matplotlib · Seaborn · HTML/CSS/JS · Chart.js · Netlify
- Lag_12 is the strongest predictor — year-over-year seasonality drives sales
- Q4 peaks every year — November 2017 hit $118,448
- Q1 is weakest — February 2014 lowest at $4,520
- 6-month forecast total: $349,126
Nikhil Varkute · Data Scientist Intern