This repository illustrates an application of the WhizML codebase for an analysis of cardiovascular disease risk.
The dataset used is the Cardiovascular Diseases Risk Prediction dataset obtained from Kaggle.
Running the eda pipeline will launch the following Auto-EDA dashboard, allowing the users to observe the dataset.
Users can implement custom functions to preprocess the data. In our case, the preprocessing codes can be found in Data_Preprocessing.ipynb, inside the notebooks directory.
The model_experimentation triggered the training of various Logistic Regression, Random Forest, and XGBoost models.
Model explainability can be further explored using the model_explainability pipeline.
Bias analysis can be performed using the Aequitas web app, with the data provided by using the bias_analysis_data_prep pipeline.
As new data is obtained, drift detection can be performed using the data_drift_analysis pipeline.
Note: To create a hypothetical example, some rows were sampled from the original dataset and were assumed to be the new data.