search

williamflaherty / pearing

KCSW project
Other
1 stars 0 forks source link

Matching Algorithm #15

Open williamflaherty opened 9 years ago

williamflaherty commented 9 years ago

I'm thinking we'll use distance to start, a la Tinda. But I'm hoping that eventually we can move to a model where we correlate #hashtags used to give a better match percentage for common interests. (This could get really interesting)

williamflaherty commented 9 years ago

For when the time comes: https://docs.djangoproject.com/en/1.7/ref/contrib/gis/tutorial/

williamflaherty commented 9 years ago

We display 5 matches at a time. There will be a way to refresh the matches. When we refresh the app will make an API call to get 5 - (# of favorited people) matches.

Initially the matches returned will have to be within a certain distance. This means we would use Location Services to get a lat/lon of a user and mark them as in that area for a set amount of time (that I assume would expire, at which time the server asks the app for an updated lat/lon?, this is less expensive than updating constantly). Then I'm thinking using GeoDjango and PostGis on the backend to handle the processing of finding other matches around them. (I considered us doing this ourselves but it's an expensive and complicated operation that I would prefer we just use something pre-rolled to do.)

Eventually I would like to do some more exciting things with the matches like matching based on frequency of tags used or getting color profiles for their pictures and matching based on frequency of colors used.