After training with the prediction model, it automatically evaluates the prediction accuracy. This screen provides a summary of the model evaluation.

The prediction accuracy is calculated by comparing the actual results with the predicted results of the prediction model that is created. We found that we can predict the complaint label with good accuracy from the number of stars at the prediction accuracy level!

Select “Precision” to see a more detailed evaluation. Evaluated values of prediction accuracy from various perspectives and tables and graphs of prediction accuracy are generated. You can scroll down to browse.

We found that the prediction accuracy is high. So why is the prediction accuracy high? Click [Contribution].

This screen allows you to see which variables are effective for the prediction and how they are effective. Predicted results of the complaint type “Contents of the post” is effective.

It also tells you what words in “Contents of the post” are likely to be of the “(a) Appearance” type.

Then click [Predict].