The author J.M. Alonso offers a conceptual method to create natural explanations to children using single classifiers.

Preparation
In order to use this method it is necessary to first obtain classifiers using WEKA. This can be done by using the following:

• J48 is the Weka class for generating C4.5 decision trees .
• RepTree is a fast implementation of C4.5 trees using information gain along with backfitting reduced-error pruning.
• RandomTree builds C4.5 trees that consider K randomly chosen attributes at each node.
• FURIA is the Weka class that implements the Fuzzy Unordered Rule Induction Algorithm.

It is then necessary to use the XAI4KIDS to determine which will be used as explanation in SCRATCH.

Within the described experiment they made use of a balanced data set.

Method
The xai4kids program for Scratch can be downloaded here . Once you have obtained your classifiers you can use this program to create explanations.

Return to the Introduction