I’m struggling with the exact same issue as described in tha last post of the discussion linked above, but unfortunately the discussion was already closed a year ago.
I would like to loop over several columns and if one value is wrong or out of a specific range ( e.g. -20 < x < 20), the applicable row should be removed only for this specific loop. Moreover, I would like to build a predictive model consisting of a predictor for each column in the same step.
As mentioned above, the linked discussion describes a similar problem but without a final solution. Any comments or recommendation are welcome!
Hi @uxsnu, in order to use the node “rule based row filter” inside a loop i use to connect the node table column to variable and connect it to the “flow variable port” of filter node. In the picture below you can see an example.
On the other hand, to be able to help you with the prediction you want to make, you must specify the problem much better, there are several prediction methods and at least the problem must be understood to know what type of prediction can be used.
I really appreciate your quick response! Please see below the first attempt of my workflow to predict the stage1 measurements or its deviations from the setpoints.
My first goal is to loop over all deviations individually and filter the deviations that are bigger than a certain value or range (e.g. -2.0 < x < 2.0) to make sure that no errors are taken into account. For example the first loop filters all rows of deviations within the applicable range and then feeds this column to the learner/predictor. The second loop filters all rows of devitions for the second measurement and then feeds this column to the learner/predictor and so on for all 15 deviations… I guess this is necessary bc every loop filters different rows and therefore feeds different data to the learner/predictor.
The final goal is to build a model that can predict the 15 deviations/measurements.
I also would like to build a workflow with multiple leaners/predictors to evaluate the results and maybe integrate a parameter optimization.
I would really appreciate your help or also of any other community member. Also feel free to message me any time (also possible via mail: uxsnu@student.kit.edu)
Hi @uxsnu, thanks for share your data flow, first of all, I think you don’t require loops, you can make a filter, they by themselves go through all the rows. On the other hand, the regression model can also be applied without the loop, and I recommend you normalize the variables before to avoid that those with higher ranges have greater weight in the prediction. Your idea is very well structured, simply eliminate the loops, apply filters, apply the learner and then the predictor, this for each of the variables to predict.
thank you again!! I have a few more questions about my workflow. Do you mind to contact me via mail (see above), so that I can ask you after your opinion how to handle several issues?? I would really appreciate it!
