Eventful

The better app for all campus events.

Comment

Inspiration

When we first started at UMass, we all had great difficulty finding events we were genuinely interested in due to the sheer number of clubs and organizations on campus, each with it's own set of poorly advertised events. Many clubs have altogether stopped using the official event listing on Campus Pulse (the official way clubs are registered) or are ineligible to use it (official university event organizers, college-affiliated clubs, etc.). In fact, most clubs resort to Instagram to market their events, which has low accessibility to students who are not involved already with that club.

Students we talked to shared a similar viewpoint: events are UMass are messy, and it's hard to find high-quality events that people are personally interested in. We were inspired by our desire to find events that we were interested in and be more active in the UMass community.

What it does

Eventful provides users with a highly personalized feed of events they would be interested in based on personal preferences, prior attendance, and detailed click analytics within the app. It also allows them to search for events the app may not have suggested to them.

How we built it

We built the application by splitting it into a few parts. The backend service, the database, the client, and the Amazon Personalize service that provides recommendations. Each team member was in charge of one of the 4 tasks and everyone helped the other team members as needed.

The events and organizations are currently scraped from Campus Pulse (UMass's official source for Registered Student Organizations), and we are adding additional sources such as University-wide events and college events like those from CICS.

Challenges we ran into

We had some issues with training the model due to a requirement of 1000 impressions. To solve the issue we generated a fake data set based on custom weightings, allowing the model to have a data set to pull from that was not entirely random.

Accomplishments that we're proud of

Getting recommendations out of Amazon Personalize to create data based on our inputs was exciting and fun. None of us had used ML prior, yet we managed to make it work.

What we learned

We learned how to integrate Amazon Personalize using user impressions to generate a more personalized experience for users. We also learned how to use React with Material UI to create a visually pleasing website.

What's next for Eventful

Better Personalization: We will need to fine-tune our model and training datasets. In addition, we are planning to include different types of recommendation like what you would see on Amazon or Netflix. These include collaborative filtering, User-User (people like you, like that), and Item-Item (if you like this, you might also like that) recommendations.

More Engaging Content

Integration with Campus Atlas: Campus Atlas is a modern maps service for campus that features indoor maps. People have a hard time finding where these events are taking place, so having step-by-step directions (indoors and outdoors) will make lives easier. In addition, there are benefits to having location-based events — say you are bored but don't want to walk too far to an event.

Share this project:

Updates