Interpreting the predictions of the model
Being able to interpret the predictions of a model helps data scientists, auditors, and business leaders understand model behavior by looking at the top important factors that drive the model's predictions. It also enables them to perform what-if analysis to validate the impact of features on predictions. The Azure Machine Learning workspace integrates with InterpretML to provide these capabilities.
InterpretML is an open source community that provides tools to perform model interpretability. The community contains a couple of projects. The most famous ones are as follows:
- Interpret and Interpret-Community repositories, which focus on interpreting models that use tabular data, such as the diabetes dataset you have been working on within this book. You are going to work with the interpret-community repository in this section.
- interpret-text extends the interpretability efforts into text classification models.
- Diverse...
