---
license: apache-2.0
library_name: scikit-learn
tags:
- tabular-regression
- sales-forecast
- gradient-boosting
- cross-sectional
datasets:
- dev02chandan/sales-forecast-dataset
metrics:
- rmse
- mae
- mape
- smape
---

# Sales Forecast Model (GBR)

**Task:** Predict `Product_Store_Sales_Total` from product and store attributes.  
**Data:** dev02chandan/sales-forecast-dataset (`raw/SuperKart.csv` with processed train/test under `processed/`).  
**Model:** GradientBoostingRegressor selected via GroupKFold CV on `Store_Id`.

## Test Metrics
- CV RMSE: 1157.1346565946897
- RMSE: 1600.05837632221
- MAE: 1405.5687461646362
- MAPE: 27.069205177956633
- sMAPE: 32.25248697544593

## Usage
```python
from huggingface_hub import hf_hub_download
import joblib, pandas as pd

pkl_path = hf_hub_download(repo_id="dev02chandan/sales-forecast-model", filename="model.pkl", repo_type="model")
model = joblib.load(pkl_path)

# X must contain the same columns used in training (one-hot is inside the pipeline)
# Example:
# X = pd.DataFrame([...])
# y_pred = model.predict(X)