Partitioning

KNIME Analytics Platform
#1

Hi everyone,

First of all sorry for my english. I'm new to knime and I'm trying to split a dataset into a training and a testing set. But In my initial dataset I have many rows belonging to the same category and I want these categories stay together after the splitting, like all the rows belonging to category 1 are in the same partition.

Little example :

Before partitioning : 

Row1 -> Value : 2 -> Category 1

Row2 -> Value : 5 -> Category 1

Row3 -> Value : 6 -> Category 1

Row4 -> Value : 3 -> Category 2

Row5 -> Value : 1 -> Category 2

Row6 -> Value : 4 -> Category 3

Row7 -> Value : 3 -> Category 3

Row8 -> Value : 1 -> Category 4

Row9 -> Value : 2 -> Category 4

 

After partitioning :

Partition 1 :

Row1 -> Value : 2 -> Category 1

Row2 -> Value : 5 -> Category 1

Row3 -> Value : 6 -> Category 1

Row8 -> Value : 1 -> Category 4

Row9 -> Value : 2 -> Category 4

 

Partition 2 :

Row4 -> Value : 3 -> Category 2

Row5 -> Value : 1 -> Category 2

Row6 -> Value : 4 -> Category 3

Row7 -> Value : 3 -> Category 3

 

Have you any idea how to do that with Knime ? Thanks in advance for your answers!

#2

I'd suggest using the Nominal Value Row Filter and then create the inverse set by using the Reference Row Filter.

#3

Hi gabriel,

Thanks for your answer.

I managed to do that by partitioning only the category randomly, and then rejoing the dataset to the corresponding with the Join node.