Case-study to show how network (or graph) analysis could help organisation of events.

Primary key: ParticipantId

Primary key: EventId

Primary key: EventId + ParticipantId

You are in the midst of organising another event. A few participants were enthusiastic, and had already signed up.

Your Agency would like to be more deliberate with attendees, so as to encourage better interactions. Specifically, it would be best if (a) remaining participants had not met participants who had already signed up, and (b) remaining participants had somethign in common with participants who had already signed up.

To elaborate on (b), it would be nice if people who haven't met each other could see each other, and say: "Hey, do you know Person XXX?" "Ya, I do know!" Not unlike how social media platforms recommend friends to us.

Use a network, since this captures social relationships. To achieve this, generate pairwise combinations of individuals who have attended the same event, and model them as a network.

Then, use metrics like Jaccard coefficient to prioritise candidates to invite to the event.

1. Organisation of datasets is key. Before we can even begin, clean tables for events and participants need to be present.
2. Running network analytics as a service is hard, but doable. It's simply software engineering.
3. Predictions should be ground-truthed. Just because Jaccard coefficient scores says they are good candidates, doesn't mean that they actually would enjoy each other's company.

DSAID has some experience in all the above! We've built and are building analytics systems with Agencies. So if you wanna chat about network analytics, reach out to [email protected] or [email protected] (author of this repo).