Multiple machine learning models for predicting major adverse cardiovascular events in dialysis with clinical and echocardiographic parameters: a retrospective cohort study

Figure 1.. Study flow diagram. Patients who met the inclusion and exclusion criteria were divided into training sets and test sets according to 7:3. Eight machine learning models were trained, and the best model was selected based on evaluation indicators and parameter optimization. SHAP value was used to analyze the effect of features on major adverse cardiovascular events in dialysis patients.

Figure 2.. Feature selection based on Lasso regression. (A) The Lasso path diagram shows the correlation between the L1 norm and different coefficients in Lasso regression. (B) The Lasso cross-validation graph shown the correlation between λ and binomial deviation.

Figure 3.. Correlation heatmap. Correlation heatmap shown correlations between the included characteristic variables.

Figure 4.. Radar charts in the training set. The radar charts show the performance of the eight machine learning algorithm models in the training set.

Figure 5.. Radar charts in the test set. The radar charts show the performance of the eight machine learning algorithm models in the test set.

Figure 6.. Model performance comparison. (A) ROC curves for different machine learning models in the training set and validation set predicted major adverse cardiovascular events. (B) Calibration curves for different models in the training set and test set. (C) DCA curves for different models in the training set and test set.

Figure 7.. Global model explanation by the SHAP method. (A) The Shap hive diagram illustrating the distribution of SHAP values for each feature. The color gradient represents the feature values (red for high values and blue for low values). (B) Characteristic effect double ring diagram, the inner ring structure representing the feature importance distribution, and the radial distance of the outer ring scatter representing the SHAP value, and the farther away from the ring the ring, the greater the impact.