Classifying different column in groups- Data manipulation

I would like to make a question. In my data, I have in every ID different features that have been measured previously. These features correspond to three different modes. In order to get my prediction model, I think it would convenient to group the different features corresponding to each mode. This way, the prediction model could be able to calculate the difference percentage of features among these three modes for every ID. Do you have any idea how can I do that?

Hi @helfortuny,
Welcome to the KNIME Forum! Can you share an example input table and the output you would expect for it? That would help understand the problem better.
Kind regards,

1 Like

Hi Alex!!
This is my data.

Every row (ID) corresponds to a person. For every person 5 vowels have been recorded in 3 modes: sitting, standing up and moving. Features from these vocals have been extracted using programming code.

The first column “f0A_median_1” corresponds to the fundamental frequency of A sitting, the second column “f0A_median_2” corresponds to the fundamental frequency of A standing up and the third column “f0A_median_3” corresponds to the fundamental frequency of A moving. The same happens with the following columns, the ones ended with “…_1” correspond to the sitting position, the ones ended with “…2” correspond to the standing up position and the ones ended with “…3” correspond to the moving positition.

Is there any possibility I can join all the columns belonging to the same position in order to find differences between same features in distinct positions?

Thank you again!

Helena Fortuny

@helfortuny I have read your description several times and I think I still don not understand what it is you are trying to do and what the data says. I read differences and models. I think you might have to do some more explaining in order for someone to help.

Maybe you can also elaborate what you have done so far. Also if it comes to models (which ones) it might make a difference how many recordings/persons you have.

There are several methods out there an din KNIME to maybe help you. I think we might need to better understand the task.

Hi Helena,
When you say that you want to join the features, how do you want to do this? Generally, many nodes that allow you to select columns have a twin-list with includes and excludes and at the top of that twin-list you can select “Wildcard/Regex Selection” instead. With that you could use a wildcard pattern like “*_1” to get all columns related to sitting, “*_2” for standing, etc. If you want to aggregate the columns by position, you could use a Column Aggregator node. If you simply want to separate the columns into different tables, you can use multiple Column Splitter nodes.
Kind regards,