This figure shows the classification accuracy for semi-amplitude predictions in the shuffled data set V1. The model uses a six-class scheme: five linearly spaced bins representing increasing planetary semi-amplitudes, along with a separate “no planet” (NP) category. The model achieves an overall accuracy of 76% for all planetary systems. Accuracy is highest for the first two lowest amplitude bins and the highest amplitude bin and decreases across the intermediate bins. The NP scenario shows notably lower accuracy, with many instances misclassified into the lowest amplitude bin.