KNIME as tool for datamining and selections

Hi Colleagues,

We are thinking of using KNIME as our main tool for datanalyses, datamining and selections of campaign. We allready know al lot of users of KNIME use KNIME mainly for their analyses and more indepth datamining.

But we would like to hear from companies who also use KNIME for selections of e-mails or DM campaigns. For selections the following functionalities are really important:

- for standard campaigns we have to be able tot standardize the workspace, so we can do the same selecties every month (without a a lot of work).

- data preparations (like cleaning/creating a field for salutation, cleaning the field housenumber, create a field which validates an emailadres).

- split files based on a specific field, or split them in random fields. 

-  deduplicate selections based on more than one variable.

-  and more ;-)

Does anyone has time to share their experiences with creating workspaces for selection campaigns.

Thanks a lot!!

Linda  (database marketeer).

 

 

Hi Linda,

as Digital Marketer I have been working with KNIME for quite some time now. I believe it has all the flexibility that is needed for your tasks and even more. Plus you can always expand it with coded snippets in R or other languages.

The advantage of KNIME is that its workflows are extremely reproducible, so you can run the exact same transformation jobs on a new dataset every week or every month and have consistency in your output.

KNIME can also be extended in many ways (via additional nodes or coding/scripting) so in case you get stuck on a task there is always a suitable work around.

I am just done on a rather large KNIME project pulling data from many different data sources, transforming and consolidating the output into a multi-level set for analysis and visualization.

Where KNIME has to grow is definitely in the data visualization part. The embedded reporting solution (BIRT) is good enough but not up-to-par with the most modern interactive tools. The visualization nodes are ok but have a retro look and feel. The new Java based nodes are going in the right direction but are not there yet.

For the rest I think it would be perfectly suited for your task, but I cannot be more precise without knowing exactly which steps your data need to go through.

Hope this helps!

Cheers,
Marco.