This figure compares the model’s classification accuracy for the monthly separated ordered validation data set M (see Sections 4.2, 6.3) across scenarios with 50, 100, and 150 observations. Accuracy consistently improves across all semi-amplitudes as the number of observations per sample increases, reflecting the expected effect of a larger observation count.