Skip to main content
Fig. 4 | BMC Medical Research Methodology

Fig. 4

From: Multivariate longitudinal data for survival analysis of cardiovascular event prediction in young adults: insights from a comparative explainable study

Fig. 4

SHAP explanation. Panel A and B: SHAP summary plot for the top-20 predictors of all subjects (A) and individual force plot of a single subject (B) for RSF model trained on trajectory clustering memberships. High feature value = cluster group 3, low feature value = cluster 1. Panel C and D: summary plot of the top-20 predictors (C) and individual force plot of the same subject (D) for RSF trained on 250 time-series extracted summary statistics. In the summary plots, each dot represents a subject. A dot’s position along the x axis (i.e., the actual SHAP value) represents the impact that feature had on the model’s output for that subject, and in this case corresponds to the survival probability (i.e., a lower SHAP value indicates a lower survival probability, or a higher CVD risk). Features are ordered along the y axis based on the mean of their absolute Shapley values. The dot’s color represents high (pink) or low (blue) value of the feature, and dots “pile up” along each feature row to show density. In the individual force plots, a longer arrow indicates a greater impact on pushing the predicted survival probability of the subject towards (pink) or away from (blue) the population average

Back to article page