Cross-validation analysis of predictions sets against target score


I have biological model based on 9 sets of theoretical predictions (each of which consist of 6 predicted values) made against the same number of  experiments which are assosiated with 9 values of scores (which are in the normalized form ranged from -1 to 1 for each system) . Within the table (see the attached picture of it) all data are sorted in the following pattern: on Y I arrange all of my 9 datasets, next 6 rows on X for each dataset are corresponded to its predicted values and last (7th) colum consisted of the experimental scoresa.  The task is to find possible correlation between predictions for each of 9 systems based on itsl predicted values (6 for each)  to corresponded experimental (score) values (1 values for each system).

Using KNIME I've succsesfully tested my prediction agains score using simple linear regression method obtaining good linear pattern between score and  prediction's values. Now I'd like to test my model using cross-validation. I've built typical crossvalidation loop consisted of  X-partitioner , Desicion Tree-learner and so on. My issue is on the step of data processing using Desicion Tree-learner which can not load data because validates Scores (last column in table) has not nominal values (it might have values -1, 0.5, 0, 0.2, 1 for [articular datasets). 
So I'd be expecially thankful for any help regarding cross-validation of such data as well as useful tutorials covering same problems.



Hi J.

our decision tree only takes categorical class columns. you would need to transform the numerical ones with a number to string node before hand. However, our Tree Ensemble Learner directly works on numerical values.

About Cross Validation, we do have one CrossValidation Workflow on our Example Server.

Let me know if I can be of any further help,

Best, Iris