Skip to main content
Fig. 5 | BMC Medical Informatics and Decision Making

Fig. 5

From: Leveraging machine learning for duration of surgery prediction in knee and hip arthroplasty – a development and validation study

Fig. 5

SHAP waterfall plot for a random patient (example) for the top 10 features. The plot must be read from bottom to top. The expected value of DOS was E[f(x)] = 46 (i.e. “blank” expected DOS without features). After adjusting for features, the expected DOS (= f(x)) was 36 for this patient (see top). Variables other than the top 10 are summarized as “217 other features”– which collectively decreased DOS by 1.8 min

Back to article page