From space analysis and predictive analytics, we see that businesses around events play a large role in its attractiveness. Therefore, combining these results with topic modeling and association rule results, event planners should take into account of the overlapping characteristics of the two groups of people: the one that attends comedy shows or arts performances and the one that are consumers to bars, or brunch/coffee shops.

Time analysis also shows that time of the day, day of the week and event durations are all important factors that determine its attractiveness. We also observe that events that happen on Monday and Wednesday mornings are more attractive than others. This interestingly leaves “cocktail” events or “church” out, as we see are popular themes from topic modeling. Therefore, we suspect that these popular events are likely to be cheap, short, fast-turnout ones such as getting coffee with a coworker.

Events with incomplete information in any of the columns that we selected are omitted from the study. If they share some common characteristics (location, type, time, etc.), or if there is a reason why these hosts/target audiences have difficulty providing the information, then they are not represented in this study.

We used equi-depth binning on some of the features, which might have separated similar data points into different bins.

Yelp event users might not be aware that their social lives are tracked and utilized to study how to sell more products/events to them. Even though the it’s unclear whether the responsibility falls on Yelp or the researchers using Yelp APIs, this is something to think about.

Data highly skewed
Attractive event definition >1.
People don’t check in on the App -> Something Yelp should consider improving

Only public transportation data is collected. No parking/taxi information.

No individual human data, can potentially study characteristics of the population that attend events and divide those into different groups and look at their different behaviors.

No attending percentage (actual attendance / registered count)

Maybe filter by business categories or topics in event descriptions, then redo cluster analysis on event attractiveness or re-test the hypotheses.

Combining Hypothesis testing, classification, association rule and topic modeling results, event planner should target venues at business centers with easy transportation options. They should specifically make sure that people don’t walk much. There’s no need to host in the most expensive locations such as Manhattan (even though it’s surrounded by many businesses), rather they should find affordable cities such as Brooklyn to lower the attendance cost to attract more participants.

Whatever their business motivation is, they should consider channeling something artsy to attract people on Thursday or Friday nights. Most importantly, find out who your audiences are and give them a reason to come to your event.