I have some questions regarding regression. I don't have wide knowledge in statistics - I'm rather technician (so sorry if questions are obvious for analysts).
I'm investigating ODM and trying to understand background of regression in ODM.
1. How regression algorithm predicts values of parameters for regression (SVM) in ODM ??
If I'm launching model in workflow it automatically calculates some values to complexity factor, svms_epsilon, std_dev, etc. How it is calculated ??
2. What is a measure of best algorithm ?? Minimal RMSE and MAE ??
If minimal RMSE and MAE, then automatic calculation gives poor results.
3. I checked several combination of parameters and concluded, that the best fit to my original data (minimal RMSE and MAE) is for gaussian svm when is:
minimal epsilon, minimal deviation and maximal complexity factor.
But I think that parameters strictly fits to input data and does not follow "random factor" (because of minimal deviation).
How to check what are optimal values for regression parameters ??
4. How to calculate "predictive confidence" of regression algorithm using ODM SQL? In workflow it is calculated automaticaly, in ODM SQL I didn't found possibility to show this measure.
Perhaps you might want to sign up for the new 2 day Instructor Led Oracle University Course on Oracle Data Mining? It helps to answer these and many similar questions. See blog entry https://blogs.oracle.com/datamining/entry/new_2_day_instructor_led
Also, the ODM Docs are posted on-line so you can always search them for more detailed answeres. [Others may respond w/ specific details to your questions. I wanted to make you aware of the great learning resource as you said that you were new to the field. Best of luck!]