Skip to main content

Table 19 Combining thermodynamics (Method 2) and N-Grams (Method 11) for training RBF-epsilon regression SVM model

From: Improving model predictions for RNA interference activities that use support vector machine regression by combining and filtering features

   train2431 train2431 train2431
   test2431 test579 test2431 10 × cross val
t- test FN2431 R MSE R MSE R MSE
0 1383(1383) 0.746 0.018 0.491 0.101 0.721 0.025
1 797(791.5) 0.755 0.017 0.480 0.093 0.709 0.026
2 443(413.3) 0.763 0.017 0.480 0.094 0.688 0.027
3 191(177.7) 0.773 0.016 0.462 0.097 0.671 0.028
4 88(75.3) 0.812 0.014 0.448 0.100 0.656 0.028
5 40(33.4) 0.818 0.013 0.442 0.097 0.659 0.028
6 18(14.8) 0.789 0.015 0.448 0.096 0.659 0.028
7 12(10.2) 0.759 0.017 0.435 0.100 0.656 0.028
8 7(5.9) 0.686 0.021 0.428 0.106 0.600 0.031
9 4(3.9) 0.587 0.026 0.335 0.111 0.586 0.032
  1. Models trained on dataset2431 and testing performed with dataset2431, dataset579 and 10 × cross validation on dataset2431, features removed by increasing stringency of t-test of individual feature to activity from dataset2431.
  2. Feature numbers in parentheses are the average number of features in cross validations.
  3. Entries in bold are column maximums and are simply visual landmarks.