The Supervised Model Dashboard allows a user to quickly assess if a dataset is suitable for modelling. This dashboard was designed to display the evaluation results of a baseline supervised machine learning model, and explain some of the model predictions. The Dashboard is run from a jupyter notebook and the display is rendered via voila.
This dashboard displays the following for a Supervised machine learning model:
Binary classification problems:
- model performance scores,
- Receiver operating characteristic curve,
- Precision-Recall curve,
- Classification Report,
- Confusion Matrix,
- Model hyperparameters,
- Sample of predicted labels with features and true labels,
- Impurity-based feature importance scores,
- Summary plot of SHapley Additive exPlanations (SHAP) values,
- Bivariate plots of Target vs Features for Top N features
Regression problems:
- model performance scores,
- Model hyperparameters,
- Sample of predicted labels with features and true labels,
- Impurity-based feature importance scores,
- Summary plot of SHapley Additive exPlanations (SHAP) values,
- Bivariate plots of Target vs Features for Top N features
Note: The numbered labels in the screenshots correspond to each widget in the custom python Dashboard class. For more details on the jupyter widgets, callback functions, process flow, please refer to the user guide or (user_guide>Supervised_Model_Dashboard.md).
The Dashboard was written in the python programming language, and runs from a Jupyter notebook. The display is rendered using voila.
After setting up the virtual environment, open the Supervised Model Dashboard notebook (notebook > Supervised_Model_Dashboard.ipynb)
Run all cells to display the dashboard.
Check out the detailed user guide or (user_guide>Supervised_Model_Dashboard.md) for details on the configuration settings, setup files, step-by-step walkthrough of all the displayed content, as well as the process flows for the classes.
See the respective Exploratory Data Analysis (EDA) notebooks (notebook > EDA) for the rationale and justification of the preprocessing steps in the CustomIngestor and CustomPreprocessor python classes.