Similar events on Gatekrash (part 1)
How I'm making the 'similar events' feature on Gatekrash as accurate as possible
This is part one of a two-part post
One of the features I've wanted to build into Gatekrash for a while now has been a recommendation engine. It would, based on the event you're looking at, recommend some similar events to you.
In the beginning I had high hopes of customising this recommendation system to each individual user based on their browsing history on the site. I've since slimmed down this ambition to a more simple solution using the data below to guide the system.
In these posts, I'll describe how the system on Gatekrash works.
What have you got to work with?
Well, I have quite a few pieces of data to work with. In no particular order, they are:
- The event title
- The time and date of each event (and the start and end times)
- The venue in which an event is being held
- The town/city in which an event is being held
- A description of the event (in some cases)
- Performers (in some cases)
- 'Also Occurring' information (how many other times an event is occurring)
As you can see, a rich set of data. Already ideas are forming about how these pieces of information can be used to influence recommendations. For example, you could argue that events happening in a few days are more useful (and therefore a better recommendation) than those events happening in a few weeks.
Likewise, you could infer that an event happening in a location closer to the user (using their current location settings are a guide) are of more relevance to the user than events happening 200 miles away.
So, what's it going to look like in the end?
The recommendation feature will exist on event pages, and at some point in the future will be made accessible via the API so that it can be embedded in listings and on the mobile version of the site (and for any developers who care to use the site).
The recommendations box will appear at the bottom of event pages, next to the 'Also Occurring' box. It'll be a simple list of events much like the 'recommended videos' on YouTube's video pages.
Then what happened?
Well, once I've launched the feature, I'll write up the whole post on how I did it. Shouldn't be too long now.