---
license: mit
---
# 🍇 Grape Firmness AutoML Model

## Model Details
- **Model type**: RandomForest Regressor (selected by AutoML search)
- **Framework**: scikit-learn
- **Preprocessing**: Custom feature engineering and scaling (`preprocess.joblib`)
- **Files**:
  - `model.joblib` — trained RandomForest model
  - `preprocess.joblib` — preprocessing pipeline (feature transforms)

---

## Task
**Regression**: Predict grape firmness (continuous value) from structured features.  
This task is designed to explore *Classical AutoML* approaches for tabular datasets.

---

## Dataset
- **Source**: [rlogh/grape-firmness-dataset](https://huggingface.co/datasets/rlogh/grape-firmness-dataset)  
- **Samples**: ~X rows, Y features (replace with actual numbers if needed)  
- **Split**: 70% train, 15% validation, 15% test  
- **Target**: `firmness` (numerical label)

---

## Training Procedure
- **AutoML approach**: grid search across models (RandomForest, SVR, etc.)
- **Search space**:
  - `RandomForestRegressor` — tuned `n_estimators`, `max_depth`
  - `SVR` — tuned `C`, `kernel`
- **Evaluation metric**: R² score (maximize)
- **Validation**: stratified split for regression

---

## Results
On the **held-out test set**:
- **RMSE**: 0.177  
- **R²**: 0.943  

The RandomForest Regressor achieved the best performance.

---

## Limitations
- **Small dataset size** → may not generalize to unseen grape varieties or measurement settings.
- **Not production-ready** → purely academic demonstration.
- **Feature engineering is minimal** → no domain knowledge incorporated.

---

## How to Use

### Install requirements
```bash
pip install scikit-learn joblib
import joblib
import numpy as np

# Load preprocess and model
preprocess = joblib.load("preprocess.joblib")
model = joblib.load("model.joblib")

# Example input: replace with real sample (shape must match training features)
X_new = np.array([[5.1, 3.5, 1.4, 0.2]])  # dummy
X_new_p = preprocess.transform(X_new)

# Predict firmness
y_pred = model.predict(X_new_p)
print("Predicted firmness:", y_pred)

## Dependencies

numpy==1.26.4

scikit-learn==1.4.2

pandas==2.2.2

joblib

@dataset{rlogh_grape_firmness_2024,
  title={Grape Firmness Dataset},
  author={rlogh},
  year={2024},
  url={https://huggingface.co/datasets/rlogh/grape-firmness-dataset}
}