Post-hoc explanation of black-box classifiers using confident itemsets

作者：

Highlights：

• Confident itemsets are used to discretize the whole model into subspaces.

• Concise instance-wise explanations approximate the behavior of the black-box.

• Class-wise explanations approximate the black-box’s behavior in different subspaces.

• Confident itemsets explanations improve the fidelity by 9.3% over other methods.

• Confident itemsets explanations improve the interpretability by 8.8%

摘要

•Confident itemsets are used to discretize the whole model into subspaces.•Concise instance-wise explanations approximate the behavior of the black-box.•Class-wise explanations approximate the black-box’s behavior in different subspaces.•Confident itemsets explanations improve the fidelity by 9.3% over other methods.•Confident itemsets explanations improve the interpretability by 8.8%

论文关键词：Explainable artificial intelligence,Machine learning,Post-hoc explanation,Confident itemsets,Interpretability,Fidelity

论文评审过程：Received 3 May 2020, Revised 28 August 2020, Accepted 28 August 2020, Available online 10 September 2020, Version of Record 19 September 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113941