Fig. 3.

Performance for the prediction of the tumor origin on the TCGA dataset. (a) The classifier model confusion matrix for TCGA test dataset and (b) TCGA-meta datasets. Rows and columns of the matrix represent the predicted classes by the model and the true classes of the tumor, respectively. The number of samples and recall are plotted upper of the matrix and precision is plotted next to the confusion matrix. (c) Overall accuracy and the weighted average of precision, recall, and F1-score of the classifier for the test dataset (left) and TCGA-meta dataset (right).