Table 2 Explainability evaluation using the TCC framework
From: Towards next-gen smart manufacturing systems: the explainability revolution
TCC categories | Sub-category | Explainability evaluation details | Scores | Overall scores | Total score | ||||
|---|---|---|---|---|---|---|---|---|---|
RFR based (S1) | XAI based (S2) | S1 | S2 | S1 | S2 | S1 | S2 | ||
Transparency | Model output interpretability | NA | Local feature contribution Global contribution Partial dependence | 0 | 0.75 | 0.22 | 0.63 | 0.14 | 0.74 |
Modelling simplicity | CS: 531,441 steps CT: 19.19 h | CS: 2000 steps CT: <1 min | 0 | 0.99 | |||||
Data transparency | Transparent | Transparent | 1 | 1 | |||||
Rule transparency | NA | NA | 0 | 0 | |||||
Model complexity | Operators: 5 Features: 4 | Operators: 4 Features: 2 | 0.1 | 0.4 | |||||
Cohesion | Explanation consistency | NA | Computational explanations | 0 | 1 | 0.22 | 0.77 | ||
Compatibility with I/O types | I/O: numeric | I/O: numeric, categorical | 0.33 | 0.66 | |||||
Integration with existing ML algorithms | Explanation support for ML types: R | Explanation support for ML types: R, C | 0.33 | 0.66 | |||||
Comprehensibility | Explanation modes | NA | Intrinsic & post-hoc | 0 | 1 | 0.00 | 0.83 | ||
Non-technical reasoning | NA | Supported | 0 | 1 | |||||
Technical & scientific insights | NA | Supported | 0 | 0.5 | |||||
