The WIT datapoint explorer and editor

In this section, we will create the datapoint editor. Then we will explain the model's predictions with the following tools:

• Datapoint editor, an interface to edit datapoints and explain the predictions
• Features, an interface to visualize the feature statistics
• Performance & Fairness, a robust set of tools to measure the accuracy and fairness of predictions

We will now create WIT and add the SHAP explainer.

Creating WIT

In this section, we will create and configure WIT.

The program first selects the number of datapoints to visualize and explore:

# @title Show model results and SHAP values in WIT
from witwidget.notebook.visualization import WitWidget, WitConfigBuilder
num_datapoints = 1000 # @param {type: "number"}

We then take the prediction labels out of the data so that the model can analyze the contributions of the features to the model:

# Column indices to strip out...