Fig. 5
From: Velo-Predictor: an ensemble learning pipeline for RNA velocity prediction
Confusion matrix of different models. Here shows the case when the number of classes d equals 4 on the DG dataset. a The performance of the XGBoost model, where each square corresponds to a predicted label and a true label for direction of RNA velocity, and the value in each square is the number of cells meeting the condition (color becomes darker as the number increases). We can see that the diagnals have darker colors which means that some samples are correctly classified. b The performance of the Random Forest model. c The performance of the Stacking model