Fig. 5: Summary of training, validation, and model interpretation pipeline. | npj Digital Medicine

Fig. 5: Summary of training, validation, and model interpretation pipeline.

From: Automatic multilabel detection of ICD10 codes in Dutch cardiology discharge letters using neural networks

Fig. 5

Data were preprocessed, vectorized, and split into a training (80%) and holdout (20%) set as shown in panel (a). Binary classifiers were trained in panel (b) and the model was evaluated in panel (c). Model interpretability was provided by using word coefficients and human interpretation of misclassification.

Back to article page