proto_framework

Basic ML frameworks for prototyping. Current examples include 1) a binary classification pipeline (RFC) for the sklearn breast cancer dataset, 2) an XGBoost time-series analysis, 3) a set of small examples such as PCA clustering and linear programming.

To run these scripts, you must create a .env file in the project root directory which will define PROJECT_ROOT, namely the full path to the directory containing this .env file.

For the RFC model

1. src/RFC/tune.py

• AUC Hyperparameter optimization for the RFC model to be trained.
• python src/modeling/tune.py --help

2. src/RFC/train.py

• Train a RFC model, with or without undersampling to address class imbalance.
• python src/modeling/train.py --help

3. src/RFC/predict.py

• Make predictions using the model trained in (2).
• python src/modeling/predict.py --help

For the time-series model

1. src/time_series/explore_ts.ipynb

• Generate a time-series dataset with mutlii-periodic signal and noise
• Illustrate extracting frequencies in the periodic data with fourier analysis.
• View the final prediction results after having run steps (2) - (4)

2. src/RFC/tune.py

• Hyperparameter optimization for the XGB model to be trained (using oldest 50% of data).

3. src/time_series/train.py

• Train an XGBoost model on this dataset, leveraging learnings from exploration and tuning in (1) + (2) (using oldest 80% of data).

4. src/time_series/predict.py

• Make predictions for the most-recent 20% of time-series data.