SES: Maximizing Social Attendance Through Strategic Event Scheduling

TL;DR

Successfully organizing a large-scale festival like Summerfest isn't just about booking the best talent—it's a high-stakes game of scheduling. This paper tackles the Social Event Scheduling (SES) problem: how to assign events to time slots to maximize total attendance while navigating user habits, resource limits, and "poaching" from competing events. The authors prove this is a hard (NP-hard) problem and offer a Greedy algorithm that outperforms standard selection strategies.

Background: Moving from Participant to Organizer

In the world of Event-based Social Networks (EBSN), researchers usually ask: "Given these events, which users should attend?" This paper flips the script. It asks: "Given these users and these competitors, when should I host my events?"

The "Satisfactory" metric here isn't just user happiness—it's the revenue and publicity generated for the organizers. To solve this, one must account for:

• Spatiotemporal Conflicts: You can't have two shows on the same stage at the same time.
• User Habits: Alice might love Pop music, but if she works every Tuesday, scheduling a concert then is a waste.
• External Competition: A rival club hosting a DJ nearby will naturally syphon off your potential crowd.

Methodology: The Logic of Choice

The authors use Luce’s Choice Theory to calculate the probability of a user $u$ attending event $e$ at time $t$. The intuition is that a user's attention is a finite resource divided among all available "attractions" (both yours and the competitors').

The Formal Core

The attendance probability ${\rho }_{u,e}^t$ is determined by the user's affinity for the event divided by the sum of affinities for all other events happening at the same time:

The Greedy (GRD) Algorithm

Because checking every possible schedule combination is computationally impossible (NP-hard), the authors developed a Greedy (GRD) approach.

1. Initialization: Calculate the potential score (expected gain) for every possible (Event, Time) pair.
2. Selection: Pick the pair that adds the most attendees to the total tally.
3. Dynamic Update: Once an event is scheduled at time $t$, the "worth" of other potential events at that same time changes (because they now compete with your own new event). The algorithm updates these scores and repeats until $k$ events are scheduled.

Note: The GRD algorithm ensures that internal competition between your own events is accounted for in every step.

Experiments & Real-World Impact

The researchers tested their model against a massive Meetup dataset from California, containing over 42,000 users.

Performance vs. Baselines

• TOP Strategy: Selecting events that are simply popular in isolation fails because it ignores that two popular events at the same time "steal" each other's audience.
• RAND Strategy: Random assignment, as expected, performs poorly.
• GRD Result: The Greedy algorithm consistently finds the "sweet spot" in the schedule, maximizing the utility (total attendees) across varying numbers of events ($k$) and time intervals ($T$).

Figure: Utility vs. number of events scheduled (k). GRD maintains a clear lead as complexity increases.

Critical Insight & Conclusion

This work highlights a fundamental truth in event planning: Context is everything. An event's success is not just a factor of its own quality, but of when it happens and who else is trying to grab the audience's attention at that exact moment.

Limitations: The model assumes user interests are static and that we have perfect knowledge of competitors' schedules. In a real-world "arms race" between venues, both parties might adjust schedules dynamically, which would require a Game Theory approach rather than simple greedy optimization.

Future Outlook: This framework is highly extensible. Tomorrow's event planning software could integrate real-time social media "hype" data into the affinity function $\mu$, allowing organizers to pivot their schedules based on emerging trends in real-time.