Blogpost #3
This week we continued to make progress on clustering posts. We’ve working on separating out data into groups, performing clustering on each of those groups, and comparing the resulting clusters.
Since last time, we’ve improved our clustering algorithm so we can now draw more meaningful insights from the clusters–– rather than simply seeing the top terms per cluster, we can also now see other information about the posts in each cluster, such as the city it was posted in it, the “type” of post (i.e. woman seeking man), or the “category” of the post, like “platonic.”
In this example, we ran the clustering algorithm on a random sample of 100,000 posts written by men and 100,000 posts written by women. Each of these groups include both ‘romantic’ and ‘platonic’ categories of posts, including varying queries such as ‘man seeking woman’, ‘man seeking man’, ‘man seeking trans’, etc.
As a result of the 2-means clustering on each of these groups, male-written posts clustered into two groups which we dubbed the "casual sex" cluster and the "romance" cluster. Female-written posts clustered into two groups which we dubbed the "friendship" cluster and the "relationship" cluster.
male poster - cluster 1
(the "casual sex" cluster)
insights
This cluster seems to represent posts written by men that are likely looking for casual sex. Words like 'travel,' 'host,' and 'discreet' indicate that the poster is looking to set up a meeting immediately. Many posts in this cluster are written by men who are seeking another man, multiple men, or trans men and women.
top terms
host, suck, cock, top, bottom, stat, clean, fuck, discreet, white, play, free, travel, nice, fun, love, prefer, email, body, open
category distribution
4014 posts tagged as "miscellaneous romance"
25 posts tagged as "strictly platonic"
top types
m4m, 0.015240618922541704
m4mm, 0.0058139534883720929
m4t, 0.0038659793814432991
m4w, 0.00062360751276978115
m4mw, 0.00060975609756097561
top cities
jacksonville, 0.010692804633548674
washingtondc, 0.010088816073122359
sfbay, 0.0089880322396808602
providence, 0.0087665143844919118
chicago, 0.0085959885386819486
denver, 0.0085023784788670793
dallas, 0.0083261058109280143
newyork, 0.0081342833585606872
male poster - cluster 2
(the "romance" cluster)
insights
This cluster contains posts written by men that may be seeking romance, love, or friendship. The most represented types in this cluster are men that are seeking one woman (m4w), multiple women (m4ww), men and women (m4mw) and trans men and women (m4t).
top terms
love, fun, friend, time, interest, seek, meet, work, nice, enjoy, white, email, open, free, thing, find, well, great, long, play
category distribution
5026 posts tagged as "miscellaneous romance"
935 posts tagged as "strictly platonic"
top types
m4ww, 0.035999999999999997
m4mw, 0.025000000000000001
m4w, 0.022336487275572161
m4t, 0.016752577319587628
m4mm, 0.011627906976744186
m4m, 0.007414983380209665
top cities
oklahomacity, 0.014692787177203919
minneapolis, 0.013712477475497737
losangeles, 0.013566938763901136
denver, 0.013314013888129477
chicago, 0.013180515759312322
sfbay, 0.012830741675486444
dallas, 0.012508432109472873
providence, 0.012470675392023707
female poster - cluster 1
(the "relationship" cluster)
insights
This cluster contains posts by women that may be seeking romance or relationship. Words like 'relationship' 'love' 'single' and 'attract' indicate that these posters are looking for more than friendship.
top terms
love, time, fun, seek, interest, meet, single, life, picture, female, relationship, work, email, thing, attract, real, enjoy, find, year, ladies
category distribution
6217 posts tagged as "miscellaneous romance"
1665 posts tagged as "strictly platonic"
top types
w4m, 0.27408056042031526
w4t, 0.23762376237623761
w4w, 0.22351871495681164
w4mm, 0.21428571428571427
w4mw, 0.19727891156462585
w4ww, 0.16666666666666666
top cities
newyork, 0.020132658963094227
jacksonville, 0.020049008687903765
seattle, 0.019932622122403144
dallas, 0.017731521634383733
sfbay, 0.017650411137344297
losangeles, 0.016126738530674933
lasvegas, 0.015303119482048263
oklahomacity, 0.014841199168892847
female poster - cluster 2
(the "friends" cluster)
insights
This cluster contains post written by women looking for friends. Many of the posts are categorized as women seeking multiple women, multiple men and women, trans men and women, or another woman. Words like 'movie' 'drink' 'friend' 'talk' indicate that these female posters are looking for friendship. The majority of posts in this cluster are tagged as "strictly platonic"
top terms
friend, hang, love, female, fun, time, meet, email, talk, interest, year, work, movie, thing, people, go, drink, hope, live, find
category distribution
804 posts tagged as "miscellaneous romance"
1314 posts tagged as "strictly platonic"
top types
w4ww, 0.14102564102564102
w4mw, 0.12730806608357628
w4t, 0.099009900990099015
w4w, 0.08774056675253826
w4mm, 0.057142857142857141
w4m, 0.044483362521891417
top cities
lasvegas, 0.0063138744716143185
jacksonville, 0.0062374693695700601
miami, 0.0056550424128180964
dallas, 0.0056471041726896022
seattle, 0.0054900492856697234
denver, 0.0045109081961834984
sfbay, 0.0044614507856386874
newyork, 0.0042701911125878345