Predicting acute kidney injury following open partial nephrectomy treatment using SAT-pruned explainable machine learning model

Table 1 Model’s features importance

Feature	Size	Renal	Age	BaseHB	Weight	Height	Creatinine	IT	IT*
Original importance	0.13	0.11	0.17	0.05	0.15	0.05	0.21	0.13	0
Estimated importance	0.14	0.11	0.19	0.06	0.12	0.03	0.23	0	0.12
Core importance	0.17	0.11	0.2	0.06	0.16	0.06	0.24	0	0

The original importance is the one obtained by a model that uses \(IT\). The estimated importance is the one obtained by estimating \(IT\) (e.g., \(I{T}^{*}\) using the other seven features). The core importance is the importance of the seven features as the weighted average in contribution to the final prediction of \(I{T}^{*}\)

ISSN: 1472-6947