"Studebaker" 😂 Dating yourself my friend. TF-IDF + K-Means + Decision Tree = Magic, though it still requires a good bit of extrapolation in order to understand the results. I find this piece to be most intimidating, and a barrier to diving right in. It seems like one must accumulate a certain threshold of experience in order to interpret the results properly. Just my 2 cents as I follow along with your videos. 👍
Interpreting cluster assignments of text documents can be particularly challenging! As you correctly point out, there's no substitute for experience. Hopefully, my next video on topic modeling using LDA will prove useful to you as an alternative as you begin your journey.
Just my brain spinning. My assumption is that I could extract a particular sets of notes in an EHR into a csv file. I can then use this process to report on particular phrases? Sorry, work has me buried so just now seeing your video and barely able to pay attention. But will revisit it soon. Hope all is well.
@@DaveOnData, I'll be re-watching this evening. Have a ticket in for the extraction since I don't have access to this particular part of Epic. I'm really looking forward to this. Hopefully will prove a theory... I'll report back with what I can.
"Studebaker" 😂 Dating yourself my friend. TF-IDF + K-Means + Decision Tree = Magic, though it still requires a good bit of extrapolation in order to understand the results. I find this piece to be most intimidating, and a barrier to diving right in. It seems like one must accumulate a certain threshold of experience in order to interpret the results properly. Just my 2 cents as I follow along with your videos. 👍
Interpreting cluster assignments of text documents can be particularly challenging! As you correctly point out, there's no substitute for experience.
Hopefully, my next video on topic modeling using LDA will prove useful to you as an alternative as you begin your journey.
Just my brain spinning. My assumption is that I could extract a particular sets of notes in an EHR into a csv file. I can then use this process to report on particular phrases?
Sorry, work has me buried so just now seeing your video and barely able to pay attention. But will revisit it soon.
Hope all is well.
@michaelt312 - Correct! For example, you can use n-grams to perform an analysis on words that frequently occur in a sequence (e.g., "united states").
@@DaveOnData, I'll be re-watching this evening. Have a ticket in for the extraction since I don't have access to this particular part of Epic. I'm really looking forward to this. Hopefully will prove a theory...
I'll report back with what I can.