A basic MLOps system that follows coding (PEP8) and engineering best practices.

The purpose of the project is to implement modular, documented, and tested software to predict which customers are most likely to churn at a bank.

This is what the directory will look like after running churn_script_logging_and_tests.py

• data
  • bank_data.csv - data used for predictions
• images
  • eda
    • churn_distribution.png - bar plot of the distribution of the target variable: Churn
    • customer_age_distribution.png - bar plot of the distribution of Age
    • heatmap.png - correlation heatmap
    • marital_status_distribution.png - bar plot of the distribution of Marital_Status
    • total_transaction_distribution.png - bar plot of the distribution of total_transactions
  • results
    • feature_importance.png - plot of feature importance for the random forest model
    • logistics_results.png - table of resulting metrics for the logistic regression model
    • rf_results.png - table of resulting metrics for the random forest model
    • roc_curve_result.png - ROC curve for the logistic and random forest models
  • logs
    • churn_library.log - logged outputs from tests in churn_scipt_logging_and_tests.py
  • models
    • logistic_model.pkl - saved logistic regression model from the train_models function in churn library
    • rfc_model.pkl - saved random forest model from the train_models function in churn library
• churn_library.py - functions for EDA, feature engineering, and model training
• churn_notebook.ipynb - jupyter notebook for doing ad-hoc analyses
• churn_script_logging_and_tests.py - tests and executes the functions in churn_library.py

Run ipython churn_script_logging_and_tests.py to test each function in churn_library.py. This should produce successfull tests in the logs folder.

joblib to save trained models

pandas for data organizing and processing

matplotlib for visualizations

seaborn for heatmap visualiation

scikit-learn for training and diagnosing the models