Figure 4.

The training process and optimal parameters of six types of full-feature ML models. (A) shows the training process of k-nearest neighbor, and the optimal K-value is 21; (B) shows the error and gamma parameter of the support vector machine during the training process, and the two optimal parameters are 0.238 and 0.1; (C) illustrates the training process of the decision tree model, and the most important branches are subarachnoid clot CT value and WBC count; (D) displays the training process of random forest, and the optima tree number of the RF model is 179; (E) shows the training process of eXtreme Gradient Boosting, and the optimal parameters are gamma of 0.25, max depth of 2, and n-rounds value of 100; (F) demonstrates the training process of artificial neural network, and generalized weights of all clinical features are seen.