tag:blogger.com,1999:blog-590043897961646114.post5089345149359330046..comments2024-02-06T02:06:06.364-08:00Comments on Engaging Market Research: Conjoint Analysis and the Strange World of All Possible Feature CombinationsUnknownnoreply@blogger.comBlogger3125tag:blogger.com,1999:blog-590043897961646114.post-31623639748829699852015-04-24T05:24:53.588-07:002015-04-24T05:24:53.588-07:00Thanks so much for the tip. I liked the stm packag...Thanks so much for the tip. I liked the stm package especially with the metadata feature!!Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-5649849166254064332015-04-23T10:37:41.082-07:002015-04-23T10:37:41.082-07:00I have suggested NMF in those cases where the data...I have suggested NMF in those cases where the data matrix is high-dimensional (lots of columns) and sparse (lots of rows with mostly zeros). I have been successful with such data matrices when there is clear separation so that we can jointly cluster the rows and columns to form blocks with higher density than other regions. Movies are good examples with the blocks representing the intersection of film genre and viewer niche. I wrote this post because I believed that the large varieties of cheeses are made and sold to different consumers for different purposes with a good deal of separation into purpose-by-cheese blocks. So, do we expect recommendation reasons to follow a similar pattern? Personally, I do not see it. Reasons for cancellation can fall into different subspaces with different customers experiencing different issues resulting from their own personal situations. Of course, if you are asking about using topic modeling for open-ended comment on a survey, then there is an R package for that called stm or structural topic models.Joel Cadwellhttps://www.blogger.com/profile/14946447393733294251noreply@blogger.comtag:blogger.com,1999:blog-590043897961646114.post-33940820284951694422015-04-23T01:47:09.927-07:002015-04-23T01:47:09.927-07:00Hi Joel do you think NMF would work on an NPS surv...Hi Joel do you think NMF would work on an NPS survey where I'm trying to understand the key topics why a customer would/would not recommend me ?Anonymousnoreply@blogger.com