great job!!
this was a big task for a small group and I'm really impressed by the careful attention to the clustering methodology and willingness to go beyond the course material/start to tackle the unsupervised learning techniques that will be covered in greater depth in DS3. a few notes:
Need to have:
- the explanation of PCA during the presentation was great; since PCA + k-means may be a more familiar clustering technique to readers than UMAP, I'd discuss some specifics of that latter clustering algorithm in the report
- I really liked the breakdown of "polarized comments" versus "nonpolarized" comments and the specific examples --- to help with interpretation, and since it's a pretty low N of comments, you may want to do automatic translation of a few examples to help with interpretation - https://www.thepythoncode.com/article/translate-text-in-python
Nice to have:
- sensitivity to 7 vote threshold
- what's the relationship between being a person who leaves a comment and being a person who votes on a comment? (know you mention join between comments and participants dataset but was less clear on what % of voters also write comments and whether the same person can leave multiple comments or only 1 comment)
- if there's time i think the next step of estimating changes over 4 time periods is interesting
- if there's time, looking at whose cluster membership is constant across the three methods of finding 2 clusters versus whose cluster membership changes
- if there's time providing some more context surrounding the introduction of uber in taiwan --- eg who was it controversial for (even tho you don't know the identity of the voters/commenters)
great job!!
this was a big task for a small group and I'm really impressed by the careful attention to the clustering methodology and willingness to go beyond the course material/start to tackle the unsupervised learning techniques that will be covered in greater depth in DS3. a few notes:
Need to have:
Nice to have: