Linear regression models


I'm a student and relatively new to Knime. I've searched everywhere for solutions to my problems and found nothing, so now I'm turning to you.

I have two-year data for energy consumption and temperature by hours. So now I have to make a linear regression model separate for each hour in each month. I've managed to make a regression with row filter nodes, but there are just too many nodes if I do so. Is there any possibility to make that kind of regression with Tree ensemble learner node? And even if I would have to do it long way, is it possible that when I import a new database with forecasted temperature, I somehow connect it with previous calculated regression coefficients by defined hour and month, so it can calculate predicted energy consumption? Or again if I do it long way, how can I put all calculated coefficients to one report or file or something, so I can import it to Excel for example.

Thank you very much for taking your time, I know that it's a lot of questions. :) I'll be very grateful for any kind of response.

I would suggest checking out the workflow examples hosted on our public server, for example the 013_XMLProcessing/013002_XMLProcessing_ExtractRegressionCoefficientsfromPMML which uses the Linear Regression Learner together with a couple of ETL nodes. Hope this helps to get started with your project and KNIME!

I would add that you can now get the coefficients directly from the node since version 2.9.  If you have many regression to run, I would investigate our looping constructs, which should allow you to set up a single regression learner node to do all of your regressions and collect the coefficients in the loop end node.  From there you can write them to an excell file or a report using BIRT.