HMM ( Supervise & Unsupervised) in RapidMiner - Hidden Markov Model

varunm1

Hello,

Hidden Markov Models are vastly used in AI, datascience and many other domains. I observed that RapidMiner doesn't have any extensions for Hidden Markove Models (Supervised & Unsupervised). It will be useful if there is an extension for classification using Hidden Markov Models.

@sgenzer@hermawan_eriadi@jpuente@luc_bartkowski

Thanks

varunm1

Any ideas on creating HMM operators in RM in the near future?

sgenzer

interesting - 10 votes for thisAny thoughts@alebo?

WT

NO easy way to do this today in RM except via R which is not ideal so would strongly support this being native in RM if possible thank you.

varunm1

@sgenzerdo you have any minimum votes that are needed to get a response on product ideas? I am sure you also look for business use cases, and the idea fit in RM, but it will also be useful if there are any periodic updates on ideas posted on the community (for example, every 6 months). Just a thought, so that we can think more about new ideas based on feasibility in the current RM structure.

sgenzer

嗨@varunm1we try to review all active Product Ideas and Product Feedback discussions at least 4x/year. Most of the time it's more often than that. There is no minimum threshold votes to guarantee that a feature will make it into the product, but I can assure you that when I hold these meetings with the product managers, we take these in order of votes.

As far as this specific idea, I just had the meeting with our Studio product manager a few weeks ago and we did discuss this. I will get some more clarity from product but my notes say that this particular feature would take some effort to implement and we are not seeing a whole lot of demand from it on the commercial side. However product agreed that it was worth keeping here and hence not declining outright.

varunm1

Thanks, Scott for informing me about the process.

sgenzer

只是另一个10美分…这是一种of thing that would be ideally implemented by the open source community as an extension (or part of one). We are working hard to make it easier for users to build their own extensions, but meanwhile everything that you would need to do this is here:https://docs.www.turtlecreekpls.com/latest/developers/extensions/

varunm1

Thanks@sgenzer, thats all needed to know. Will take a look at the link you shared.

MartinLiebig

To add, First-Level HMMs are implemented in SMILE (https://haifengl.github.io/smile/api/java/smile/sequence/HMM.html), which is 'reasonably straight-forward' to add to the SMILE extension. Would this be what you search for,@varunm1?

BR,

Martin

varunm1

Hello@mschmitz

This does it. Once you were able to add it, I will try it in RM.

Thanks
Varun