For all error metrics, the SVM is the best choice. The differences are higher for
small tolerances (e.g. for T = 0.25, the SVM accuracy is almost two times better
when compared to other methods). This effect is clearly visible when plotting the
full REC curves (Fig. 3). The Kappa statistic [26] measures the accuracy when
compared with a random classifier (which presents a Kappa value of 0%). The
higher the statistic, the more accurate the result. The most practical tolerance
values are T = 0.5 and T = 1.0. The former tolerance rounds the regression
response into the nearest class, while the latter accepts a response that is correct
within one of the two closest classes (e.g. a 3.1 value can be interpreted as grade
3 or 4 but not 2 or 5). For T = 0.5, the SVM accuracy improvement is 11.7