What makes classification trees comprehensible?

作者：

Highlights：

• In-depth survey for empirical study of the classification-tree comprehensibility.

• Objective measurements suggest the most influential parameter: the depth of leaves.

• Number of leaves is a relevant comprehensibility measure only for complex trees.

• Tree visualization style and layout significantly influence the comprehensibility.

• Proposed 2 comprehensibility measures considering semantics and structure of the tree.

摘要

•In-depth survey for empirical study of the classification-tree comprehensibility.•Objective measurements suggest the most influential parameter: the depth of leaves.•Number of leaves is a relevant comprehensibility measure only for complex trees.•Tree visualization style and layout significantly influence the comprehensibility.•Proposed 2 comprehensibility measures considering semantics and structure of the tree.

论文关键词：Classification tree,Comprehensibility,Understandability,Interpretability,End-user survey

论文评审过程：Received 13 January 2016, Revised 7 June 2016, Accepted 7 June 2016, Available online 16 June 2016, Version of Record 25 June 2016.

论文官网地址：https://doi.org/10.1016/j.eswa.2016.06.009