Porting from Orange3 to DataRobot

[hcchen]

We are trying to develop an application that finds similar bug-records from our bug database so as to to help engineers to fix them by leveraging similar experiences.

It's not a suitable way If search the database directly by key words because too many will be found in one hand, if provide more key words in another hand then too few or nothing will be found. And the results are not accurate after all.

The result can be good if we tried to see an *entire bug record* as a vector point in space and find its nearest neighbors in the space. This experiment is done by using Orange3 data mining to vectorize the bug-records into Bag of Words array and then get Distances between each other and then get nearest neighbors of a certain bug. 

We'd like to porting the solution to DataRobot but need some hints for how to do that? We can't find key words like "BoW" or "Distance" in https://datarobot-public-api-client.readthedocs-hosted.com/en/v2.25.0/index.html  though. And we need more directions to this Q&A.

Best regards!

H.C. Chen

Wistron Corp.

[Dan_Becker · Accepted solution]

Great timing to bring this up @hcchen !

There are a couple approaches we support today that might be a good fit (or might not, depending on some details).

We're also in the early stages of a feature that's more narrowly designed to show the nearest neighbors in vector space. Are you open to a call? I'd love to talk through your use case to figure out

1) How we can help today, and

2) What we can do in the nearest neighbors feature we develop to potentially make it a better fit for your use case when it's ready?

If you are open to chat, drop me a line at dan.becker@datarobot.com and we can schedule it.

Dan Becker

VP, Decision Intelligence at DataRobot