https://github.com/suredream/ag-seed-prod
https://github.com/suredream/ag-seed-prod
Last synced: about 1 year ago
JSON representation
- Host: GitHub
- URL: https://github.com/suredream/ag-seed-prod
- Owner: suredream
- Created: 2025-06-06T05:39:20.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-06-09T20:54:07.000Z (about 1 year ago)
- Last Synced: 2025-06-09T21:18:46.422Z (about 1 year ago)
- Language: Jupyter Notebook
- Size: 419 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
- Author: Jun Xiong
- Date: 2023-06-09
# ag-seed-prod
## Overview
This project is for Corteva Agriscience Case Study, which aims to predict predicting product units using machine learning models and showcase the intergration of genAI. It includes modules for data processing, model training, and a dashboard for visualization and interaction. The project proposal a XGBoost tree approach + residual correction layer as a PoC.
The codebase is built for PoC purpose only. It is not intended for production use. Check the `Todo` section for more details.
## File Structure
The project is organized as follows:
- [`README.md`](README.md): This file, providing an overview of the project.
- `config/`: Contains configuration files (TOML format) for different models and pipelines.
- [`residual.toml`](config/residual.toml): Configuration for the residual model.
- [`xgb.toml`](config/xgb.toml): Configuration for the XGBoost model.
- `src/`: Contains the source code for the project.
- [`__init__.py`](src/__init__.py): Initializes the `src` directory as a Python package.
- `utils.py`: Includes utility functions for model loading, evaluation, and confidence interval calculation.
- `dashboard/`: Contains code for the Streamlit dashboard.
- [`app.py`](src/dashboard/app.py): Main application file for the dashboard.
- `pipelines/`: Contains code for the data processing and model training pipelines.
- [`xgb.py`](src/pipelines/xgb.py): XGBoost pipeline.
- [`residue.py`](src/pipelines/residue.py): Residual pipeline.
## Model Performance
```
xgboost[after grid search]
Metric Train Test
MSE 16.2983 37.7282
MAE 2.2088 3.4257
R2 0.8131 0.6748
residual[dummy model]
Metric Train Test
MSE 16.8397 35.8759
MAE 2.3359 3.3777
R2 0.8069 0.6908
```
## Installation
enable you have `uv` with `python 3.10` in your environment.
```bash
uv init
uv sync
```
## Usage
1. Run the model training pipeline:
```bash
uv run model_update.py --model residual
```
2. Run the main.py to start the api server:
```bash
PYTHONPATH=src uv run uvicorn main:app --reload --port 8000
```
3. Run the Streamlit dashboard:
```bash
uv run streamlit run src/dashboard/app.py
```
4. Access the FastAPI documentation: [http://localhost:8000/docs](http://localhost:8000/docs)
5. Access the Streamlit Dashboard: [http://localhost:8501](http://localhost:8501)
## Running Unit Tests
To run the unit tests, use the following command:
```bash
uv run pytest
## TODO
- Apply Log (Units+1) conversation to make the prediction more robust
- Scenarios analysis
- Fix the predict API
- Clean up the code; docstrings
- Add more detailed documentation
- Unit test
- Dockerization for deployment