Explaining Black-Box Models for Biomedical Text Classification

In this paper, we propose a novel method named Biomedical Confident Itemsets Explanation (BioCIE), aiming at post-hoc explanation of black-box machine learning models for biomedical text classification. Using sources domain knowledge and confident itemset mining method, BioCIE discretizes the decision space into smaller subspaces extracts semantic relationships between input class labels in different subspaces. itemsets discover how concepts are related to black-box's space. uses approximate behavior individual predictions. Optimizing fidelity, interpretability, coverage measures, produces class-wise explanations that represent boundaries black-box. Results evaluations on various classification tasks demonstrated can outperform perturbation-based set methods terms producing concise, accurate, interpretable explanations. improved fidelity instance-wise by 11.6% 7.5%, respectively. It also interpretability 8%. be effectively used explain model semantically relates texts labels. The source code supplementary material available https://github.com/mmoradi-iut/BioCIE.