Find original data point from classification learner scatter plot

I want to find which row in my original data set a particular point in the scatter plot of the classification learner corresponds to, as I have additional labels etc. there that will help me understand which factors affect the ability of my model. Using a data tip isn't enough as only tells me the true/predicted class and the predictors that are being plotted.

Answer 1

[Drew]

In the R2023b release of Classification Learner, the row of the dataset can be determined from one of the "Select Query Point" scatter plots that are available on the Explain tab:

• Go to the "Explain" tab
• Choose "Local Shapley" or "LIME" on the toolstrip. These are local interpretability techniques which require the selection of a query point. A query point is one observation (one row) from the dataset.
• The datatip on the "Select Query Point" scatter plot shows the index (row) of the data. By default, the datatip for index 1 is shown. Right-click on datatips to delete them. Select another data point to bring up another datatip. The controls for the scatter plot are underneath the plot.

• To create more space for these two plots, the model gallery and the control panel can be minimized, as seen in the screenshot below.

• The corresponding row of the data table can also be seen with the "Table" option. In the table, notice the highlighting of the currently active query point (data row). The table view enables inspection of the values of other predictors, and sorting by the column of choice.

If this answer helps you, please remember to accept the answer.