Figure 11
The four panels arranged in a two-by-two grid are labeled “(a)”, “(b)”, “(c)”, and “(d)”, each showing a line graph illustrating the sensitivity of different parameters on model accuracy. In all panels, the vertical axis is labeled “Accuracy” and ranges from approximately 0.88 to 0.98 in increments of 0.02 in panels “(a)” and “(b)”, from 0.80 to 1.00, with the intermediate markings at 0.83, 0.85, 0.88, 0.90, 0.93, 0.95, and 0.98 in panel (c), and from 0.88 to 1.00 in increments of 0.02 in panel “(d)”. In panel “(a)” titled “Sensitivity of K in K N N”, the horizontal axis is labeled “K in K N N” and ranges from 1 to 10 in increments of 1 unit. The plotted points fluctuate around 0.90 to 0.96, increasing from about 0.90 at K equals 1 to around 0.945 at K equals 2, decreasing slightly at K equals 3, rising again and reaching the highest value near 0.96 at K equals 5, then dropping to about 0.90 at K equals 7 before gradually increasing toward approximately 0.94 at K equals 10. In panel “(b)” titled “Sensitivity of Distance Metric in K N N”, the horizontal axis is labeled “Distance Metric” and includes three categorical values: “Euclidean”, “Manhattan”, and “Chebyshev”. The plotted values show the highest accuracy near 0.964 for Euclidean, slightly lower near 0.943 for Manhattan, and the lowest around 0.91 for Chebyshev, indicating a decreasing trend. In panel “(c)” titled “Sensitivity of Learning Rate”, the horizontal axis is labeled “Learning Rate” and includes values 10 to the negative 5 power, 5 times 10 to the negative 5 power, 10 to the negative 4 power, 5 times 10 to the negative 4 power, and 10 to the negative 3 power. The plotted accuracy rises from approximately 0.85 at 10 to the negative 5 power to a peak around 0.96 at 10 to the negative 4 power, then decreases to about 0.91 at 5 times 10 to the negative 4 power and further to roughly 0.85 at 10 to the negative 3 power. In panel “(d)” titled “Sensitivity of Dropout”, the horizontal axis is labeled “Dropout Rate” and ranges from 0.1 to 0.5 in increments of 0.1. The plotted values increase from approximately 0.913 at 0.1 to about 0.96 at 0.3, then decrease to around 0.93 at 0.4 before slightly increasing again to near 0.94 at 0.5. Note: All numerical data values are approximated.Accuracy comparison of different hyperparameters. (a) Sensitivity of K in KNN. (b) Effect of distance metric on KNN. (c) Sensitivity of learning rate. (d) Sensitivity of dropout. Source(s): Figure created by authors