Open datasets : 40,000 de-identified critical care patients


(Jasper Blues) #1

If you're interested in the potential of graph for healthcare applications, here's a data set to explore:

MIMIC is an openly available dataset developed by the MIT Lab for Computational Physiology, comprising de-identified health data associated with ~40,000 critical care patients. It includes demographics, vital signs, laboratory tests, medications, and more.

(moderators: perhaps a datasets channel/tag would be useful?)


(Mike Oaten) #2

Yes, agree a dataset channel or tag would be a great addition. Particularly useful to highlight any that have been natively built on or converted to a graph data model = 'graph-ready'?

We're working on one for UK company and company officers UK company and officer data neo DB access - anyone interested?


(Jasper Blues) #3

Here is another dataset:

http://jmcauley.ucsd.edu/data/amazon/

It contains product reviews and metadata from Amazon, including 142.8 million reviews spanning May 1996 — July 2014.


(Jasper Blues) #4

And here is a repository of recommender system datasets: https://cseweb.ucsd.edu/~jmcauley/datasets.html


(Karin Wolok) #5

Thats a really good suggestion. My only concern would be congestion of too many categories. We are currently working on a new UI for the Community Site that will help navigate a bit.
Thinking of if we should make it a category or a tag, where would you suggest it be placed under? @jasperblues any suggestions? @michael.hunger?


(Karin Wolok) #6

Maybe we put data sets as a tag under Projects category?


(Jasper Blues) #7

I'm not sure how the new UI will look, however this is cool!

Do I understand the model correctly:

  • Categories are the top level menu, and all content falls into a broad category? We can discover it through a newsfeed type interface with new and high ranked content bubbling to the top. (I agree that too many of these would be overwhelming).

Meanwhile tags are something that can be associated with content in various categories?

Perhaps it would be cool if folks could:

  • browse & search tags <-- maybe using concept matching here. Use them as a process to discover relevant content. Possible UI could present an actual tag cloud, eg https://bl.ocks.org/jyucsiro/767539a876836e920e38bc80d2031ba7
  • Having done that: choose tags that they want to follow or be notified about. Then if someone posts a new piece of content with the dataset tag, I get a notification.

Also would be cool to discover like-minded colleagues by matching the tags that they post under.

Another option: categories can be followed/unfollowed, and start with a default set.

#random ideas - not sure if feasible to implement ;)


(Karin Wolok) #8

I'd actually love to show you the layout if you're interested. I'd love your feedback! Maybe you can help me brainstorm about this after I share our plans with you. :)


(Jasper Blues) #9

Sure, would you like to do that over a realtime channel? (video chat w/ screen-sharing, I mean) ? Maybe we could use neo4j-users slack channel, since this feature isn't in community.neo4j.com (yet!?)


(Jasper Blues) #10

^-- CLEVR graph aims to help further research into machine reasoning on graph datasets. It contains a set of questions and answers about transport network graphs.