Interface for KNIME
MuehliMan
New Altair Community Member
Hi,
In the last weeks got to use KNIME as pipelining software for some of my problems. As there are already interfaces to import functions from Weka and R, I wanted to ask if a similar interface is planned for RapidMiner components?
Best wishes,
Markus
In the last weeks got to use KNIME as pipelining software for some of my problems. As there are already interfaces to import functions from Weka and R, I wanted to ask if a similar interface is planned for RapidMiner components?
Best wishes,
Markus
Tagged:
0
Answers
-
Hi,
although I am not one of the RapidMiner developers I do not think that this is very likely since RapidMiner is licensed under a real open source license (namely a GPL variant) and Knime is not - at least not in the original sense. But this is not only a political issue for some hardcore nerds but a real legal one: it is simply not allowed to embed a GPL tool (like Weka - what already has been done! - or RapidMiner) into a non-GPL tool (like Knime).
Of course this might not be an issue for many users, however, for our company at least this fact definitely rules out Knime as a data analysis tool since we are not going to trust a software vendor who does not obey the license statements of the software libraries they embed or connect to. So we decided to go for RapidMiner and this actually turned out to be exactly the right decision: after visiting a training course (greetings to Ralf if he reads this) I quickly learned myself that in my opinion RapidMiner is even a better data transformation and integration tool than Knime - even if it does not stick the "pipelining" label to itself. However, in contrast to the legal issues stated before, the second argument is of course only my own opinion and would of course not principally rule out an integration of RapidMiner and Knime.
Sorry, this post might sound a bit harsh but I sort of got really frustrated after working for several weeks with Knime and then running into those legal issues which are simply not acceptable for us and I am sure also not for many others.
Kind regards,
Harun0 -
Hi,
Well I am not an expert in legal issues. I just use tools and as Weka, R and some other Platforms I happen to use are implemented (at least some our their functionalities) I just wanted to know this could be the case with RM too. I was not aware that there is a legal issue. I cannot comment your personal attitude towards KNIME as I have been using it not that long.
But after all thank you for your information.
Best wishes,
Markus
0 -
Hello ...
Another point of view: If RapidMiner were accessible via KNime, KNime would keep the primary focus as uber-integration tool and hence would steal credit from RapidMiner. I bet that if you ask in the KNime forum for a RapidMiner plugin encapsulating KNime, you would recieve a similar result.
just my cents...
regards,
Steffen0 -
Hi,
and also a note of an "official" guy ;-)
The licensing problem is indeed an issue and would prevent the integration of RapidMiner into KNIME. As an open source provider, I actually share Harun's point of view and cannot see how it is legally possible that Weka is integrated into KNIME at all without KNIME being released under the (A)GPL. But this is not my concern but that of KNIME and Pentaho.
For all others, who read this far and want to hear something more interesting than legal issues, I have a (hopefully) nice surprise. Namely a first screenshot of the current state of the new graph based process design of RapidMiner 5.0. In the next version of RapidMiner, it will be possible to use a data flow design like that known from Clementine (and also KNIME, hence I post this in this thread) as well as the operator tree based design known from previous versions of RapidMiner. Here is a small glimpse:
[img width=671 height=248]http://rapid-i.com/images/stories/rapidi/yale/concepts/graph_layout_first.jpg[/img]
Please note that this is not the final look of the data flow visualization but you should at least get the feeling what will be possible with RapidMiner 5.0. And as you can see, keeping both options (the tree as well as the flow) fills the gap between the efficiency of the operator tree (drawing all those arrows takes soooo much time ) and the easiness of a flow / graph based layout when it comes to visualizing and understanding the data flows.
Cheers,
Ingo
0 -
now we're talking! looks great, would love to see this in v5!0
-
Hi,
thanks
The graph layout is really a nice addition to RapidMiner and will probably ease process design a lot - especially for beginners. Myself, I will probably also work with the operator tree since if you are used to it things are much faster. However, I also really like the explicit visualization of the data flows and the overview over more complex processes so probably also more experiences users will appreciate it.
We will demonstrate some new aspects of RapidMiner 5 from time to time during the next weeks / months and we really eager to get your comments on those.
Cheers,
Ingo0 -
Hello,
well I'm not an expert in any topic but I have been in contac with one of the Knime developers,
according to implement Knime code in Java applications (which would be necessary here).
As far as I unterstood: Knime is really heavyly interlinked with Eclipse. To load and/or use Knime
workflows it should at least only consist of Weka modules. Otherwise the RapidMiner developers
hat to re-develop all Knime data mining components (which I bat they don't like to do ). Otherwise
one could use Knime via a shell command which neither isn't comfortable nor makes sense.
Finally I think there is only little what you can do with Knime but not with RapidMiner. I just became
accustomed to the philosophy of RapidMiner workflows, nevertheless, the network type of Knime
workflows seems to give more certainty for the user to see and understand what happens. Knime
does this quite well and intuitive while Weka for instance did realize it terrible.
Finally for the data miner the power of components is much more important than the GUI. Especially
in the field of meta methods and visualization (in my opinion) RapidMiner is far ahead of Knime.
On the other hand Knime is more advanced in data manipulation (again only my opinion).
Therefore we all should use forums like this to help to improve the software in the way we like
and need it !!!
0