I work for a medium-sized non-profit. We have a database with approximately 40,000 profiles of folks (mostly patrons). For each show a patron attends, we add an attribute that indicated that they're into that kind of show. For example, if they saw a folk music show, they would get a Folk Music tag.

Does it make sense to use some form of the K-nearest neighbors algorithm to figure out who we could target to purchase tickets? Maybe along with some location data?

For instance, we have about a 100 seat theater, so if we have sold out about half a show, would it be effective to use KNN to try to find other patrons similar to those 50, or is that too small a sample-size?


This is a classification problem that tries to answer the question: who is likely to buy a ticket to a particular type of show. If you have information about "the previous shows" that each patron attended, you can absolutely use KNN to find which patrons have similar aggregate viewership to a particular type of show.

This might not work if you just have demographic information because there will be a lot of people that share the same demographics. However, as you mentioned, you can use the lat-long or zip code level information to get US census demographic data.

Evaluating the accuracy of this model is a little bit tricky, since you don't know who didn't attend a show due to external factors, but maybe you just want a quick and fast heuristic (which in my opinion sometimes works best).

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.