Figure 5

Regression metrics. These are the MSE and \(R^{2}\) values for predicting the post-move means of the number of calls ((a) and (b)) and the fraction of calls ((c) and (d)) using the log-transformed pre-move counts ((a) and (b)) and the pre-move fractions ((c) and (d)). The \(R^{2}\) for the SVR with a polynomial kernel in predicting the post-move counts is negative, while the corresponding MSE (\(1.3\times 10^{7}\)) is much higher than the MSE of the other models