Skip to main content
Fig. 5 | Perioperative Medicine

Fig. 5

From: Can a deep learning model based on intraoperative time-series monitoring data predict post-hysterectomy quality of recovery?

Fig. 5

SHAP summary plot and feature ranking. SHAP values for the twenty most important features used in the logistic regression model (a, b) and random forest model (c, d) are shown. In plots a and c, each point represents a specific feature’s SHAP value in an individual patient. In plots b and d, a specific feature’s absolute SHAP values for all patients were averaged. The larger a feature’s absolute SHAP value, the larger the impact of the feature on patient’s outcome. A positive and negative SHAP value corresponds to a higher and lower likelihood of having an unsatisfactory outcome, respectively. The mean absolute SHAP value of all patients reflects the significance of the feature in driving model’s prediction, i.e., the higher the mean, the more significant the feature for prediction and vice versa. In plots a and c, the actual value of the feature for each patient is color-coded, with red color representing higher values and blue color representing lower values. Of note, a specific feature’s SHAP value and actual value are different. SHAP SHapley Additive exPlanation, RR respiratory rate, DBP diastolic blood pressure, EtCO2 end-tidal carbon dioxide, SBP systolic blood pressure, MAP mean arterial pressure, SD standard deviation

Back to article page
\