amesml¶
Reproducible Ames Housing price prediction as an installable package, a command-line tool and a REST API.
amesml turns a university house-price notebook into software you can install, test, version and deploy. It predicts the sale price of a house in Ames, Iowa from 79 property features, and separately whether it sells at or above the median price.
pip install "amesml @ git+https://github.com/Ahsan-muhd-444/house-price-prediction@v0.1.1"
amesml train --model gbm
Why amesml¶
- Reproducible. The default models reproduce the original notebook's published results closely - within about 5% - every model is seeded, and each saved model records the exact configuration, library versions and a fingerprint of the data it was trained on.
- Installable. A typed Python package with a CLI, tested on Linux, macOS and Windows across Python 3.9-3.13, with 99% test coverage.
- Serveable. A FastAPI service with validated requests and an OpenAPI page, packaged as a Docker image that starts with trained models.
Results¶
On the notebook's held-out split (292 houses):
| Model | MAE | RMSE | R² |
|---|---|---|---|
| Linear regression (the notebook) | $21,140 | $31,791 | 0.868 |
| Gradient boosting | $16,826 | $29,462 | 0.887 |
| Model | Accuracy | F1 | ROC-AUC |
|---|---|---|---|
| Logistic regression (the notebook) | 94.5% | 93.9% | 0.982 |
| Gradient boosting | 95.2% | 94.8% | 0.989 |
All models, with cross-validation - Model card
Where next¶
- Install and work through the 5-minute quickstart.
- The user guide covers the CLI, the Python API, the configuration file, the REST API and Docker.
- Architecture and the design decisions explain how it is built and why.