Setup

Python Environment

Environment is using conda. To install conda, follow the instructions here. You can then create a new environment using the following command:

conda env create --prefix ./.conda --file environment.yaml

where <env> is the name of the environment you want to create. You can then activate the environment using:

conda activate ./.conda

To export the environment (to add new packages), use:

conda env export --file environment.yaml

Notebooks

The notebooks are located in the notebooks directory:

• notebooks/h1st_model.ipynb: My version of https://www.kaggle.com/code/wadjihbencheikh/mod-le-1-ms-azure. It requires csv files from Kaggle
• notebooks/Aitomatic.ipynb: Final solution, using provided data files.

Training code

Main executables are train.py and predict.py. In future, they would include CLI and API. Right now, they are just executable scripts. The model logic is defined in predictive_maintenance/feature_transformers.py and predictive_maintenance/modeler.py. They make use of the generic implementation of XGBoost model (as part of H1st framework) is in ml_model/ml_xgboost.py.

Application

predict.py is but a demo/test script. The real application is in streamlit_app.py. It is a Streamlit app that uses the same model as predict.py to predict machine failures and visualize the results. To run the app locally, use:

streamlit run streamlit_app.py

Streamlit cloud will automatically updat the app at https://marov-predictive-maintenance-model-streamlit-app-0mbk0u.streamlit.app when you push to the repo.