ABSTRACT

One common Machine Learning (ML) task is to train a model to classify a data point. The model users may want to know why a model made the prediction that it did. That is, they want to be able to interpret the model or see explanations for its predictions. One particular kind of explanation is a counterfactual, which shows how a data point can be changed so that the model assigns it a different class label.

In this project, we will develop a tool that lets users explore counterfactual explanations for a given dataset and ML model. This tool allows the user to select an instance from their dataset, generate counterfactuals for that instance, and visually compare the instance with the counterfactuals.