SUMOhunt: Combining Spatial Staging between Lysine and SUMO with Random Forests to Predict SUMOylation
Figure 2
Comparison of accuracy measures (-axis) at several -fold cross-validations, LOOCV and self-consistency test (-axis). In -fold cross-validation, after dividing dataset into sections, typical training procedure is conducted using sections while the remaining one is used as test. This is repeated times until every set has been used as test exactly once. For eight -fold cross-validations performed, the value of was kept from 3 to 10 having average AC at 81%, SN at 83%, SP at 79%, and MCC at 0.66. LOOCV is a type of -fold cross-validation in which is equal to the number of total instances. It has the similar average as -fold. Self-consistency is a type of test in which prediction of every instance is done using the rules of the training dataset itself. This is done for each and every instance in the training dataset. In this test, 100% result is achieved for all accuracy measures.
