Within both groups there are algorithms with significantly different test errors (without info: p-value: 0.0103, with info: p-value: 6.467·10-6⁠, Friedman's rank sum test with Davenport correction). Pairwise comparison of results without info shows that GPC is better than GP (p-value: 0.011) and AML (p-value: 0.043) using Friedman's rank sum test with Bergman correction. For the problem instances with noise, AML produced the best result for only 1 out of 194 instances. For the instances without noise, the results of AML are similar to results of GP, GPC, and ITEA. Pairwise comparison of results with info shows that SCPR is better than the other algorithms and no statistically significant difference was found between GP, GPC and FIIT. The p-values for all pairwise tests are shown in Table 5.

Table 5:

p-values for pairwise comparison of algorithms in both groups (Friedman's rank sum test with Bergman correction).

without infowith info
GPGPCITEAAMLGPGPCFIITSCPR
n/a 0.011 0.325 0.499 n/a 0.151 0.353 0.002 
0.011 n/a 0.325 0.043 0.151 n/a 0.054 0.000 
0.325 0.325 n/a 0.353 0.353 0.054 n/a 0.028 
0.499 0.043 0.353 n/a 0.002 0.000 0.028 n/a 
without infowith info
GPGPCITEAAMLGPGPCFIITSCPR
n/a 0.011 0.325 0.499 n/a 0.151 0.353 0.002 
0.011 n/a 0.325 0.043 0.151 n/a 0.054 0.000 
0.325 0.325 n/a 0.353 0.353 0.054 n/a 0.028 
0.499 0.043 0.353 n/a 0.002 0.000 0.028 n/a 
View Large

Close Modal
This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal
This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy. No content on this site may be used to train artificial intelligence systems without permission in writing from the MIT Press.
Accept