Roskilde Festival Artist Audience Overlaps (2017)

The predictive case study on the Roskilde festival stems from an unpublished research project in which our research group volunteered to perform data analytics at the 2017 Roskilde Festival. For this case study, the Facebook walls of several hundred artists that were scheduled to play at Roskilde Festival 2017 were both fetched and analyzed through use of the Social Set Visualizer.

The Roskilde Festival case study has two major goals. First, it is to investigate if there is a way to optimize the scheduling of artists on different stages with a special focus on crowd safety. This is needed, because critical crowd safety situations can potentially arise as large crowds move from one artist to the next one which plays at a different stage. The goal is to investigate if these critical crowd safety situations can be predicted by crowd movements based on social media datafrom Facebook.

The second goal is to predict the largest concert at the festival. For this, the Social Set Visualizer software tool is used to quantify audience overlaps between the Facebook audience of Roskilde Festival itself and each of the scheduled artists.

The cardinality of each set intersection is quantified, and a prediction of the TOP 10 artists is presented.

Prediction of Crowd Movement

With regard to the first objective, the Social Set Visualizer enables prediction of crowd movement. Thus, actionable findings with special relevance for crowd safety efforts at Roskilde Festival have been generated through use of the Social Set Visualizer software tool.

The methodology is based on information of artists who are scheduled in sequence to each other on separate stages. They are seen as potential source of massive crowd movement between concert stages and therefore as a potential threat and hazard to crowd safety. For this, we apply the Social Set Analysis methodology to the Facebook data fetched through the built-in crawler of the Social Set Visualizer.

Hence, we can quantify shared audiences between scheduled artists and highlight potential crowd movement safety risks well in advance. For this purpose, the artist Facebook audiences were intersected first with the Roskilde Festival Facebook wall, in order to identify fans of each artist who are likely to go to Roskilde Festival 2017. Subsequently, these artist and Roskilde Festival set intersections were again intersected with each other with a special focus on artists with neighboring time slots.

The cardinalities of these inter-artist set intersections are calculated and visu-alized with the Social Set Visualizer software tool. Afterwards, the calculated car-dinality information can be overlayed on top of the festival programme. Figure 5.6 showcases the scheduled concerts and the cardinality of each pairwise intersection for the festival on Friday, 30 June 2017. This case study depicts thefirst time crowd movements in a music festival are predicted through the use of Big Social Data Analytics.

76 Chapter 5. Evaluation

Figure 5.6: Overlaps between Facebook audiences of different artists at Roskilde Festival 2017

The detailed predictions for potentially safety-critical large-scale crowd move-ments during the 2017 Roskilde Festival were as follows:

(1) For Thursday, 29 June 2017, the concert of Elza Soares at Avalon stage with Solange immediately following at Arena stage was identified as highest risk for massive crowd movement of 2,485 festivalgoers between stages based on Big Social Data.

(2) ForFriday, 30 June 2017, the concert of Kano at Apollo stage with Foo Fighers at the same time at Orange stage was identified as highest risk for massive crowd movement of 1,554 festivalgoers.

(3) ForSaturday, 1 July 2017, the concert of Anthrax at Arena stage with Arcade Fire immediately following at Orange stage was identified as highest risk for massive crowd movement of 5,111 festivalgoers. Further movements from one stage to the other are expected in case festivalgoers decide that the Arcade Fire concert will be more entertaining.

Due to a lack of data from the festival organizer, an exhaustive validation of the stated predictions is not possible. Official audience measurements are required in order to perform a thorough comparison of the prediction with actual numbers. This predictive case study focuses on the utility of the Social Set Visualizer in a festival analytics scenario. It highlights that the set-based approach to Big Data Analytics is applicable to a variety of real-world data science research questions.

5.2. Predictive Case Studies 77

Prediction of Concert Attendance

The second goal of this case study is to predict concert attendance based on social media data from Facebook. This prediction uses the simple heuristic of pairwise intersections between the Roskilde Festival Facebook page and each artist. The cardinality of these set interactions is calculated in order to identify the artists which are most popular with the Facebook audience of Roskilde Festival.

The underlying assumption of this approach to predicting concert attendance is that the social media audience on the Roskilde Festival Facebook page is represen-tative of the actual on-site audience of festivalgoers. If this assumption holds, the social media interactions with both the Facebook pages of the artists and of Roskilde Festival carry some sort of signal that lets us predict a relative ranking of how many people will show up at the actual concert during the festival.

Figure 5.7 showcases the prediction of concert attendance at Roskilde Festival 2017 based on set-based artist overlaps with the official Roskilde Festival Facebook

RF 2017 Concert Attendance Prediction via Social Set Analysis of Facebook Audience

27 Jun 2017 by Benjamin

# Artist # FB overlap w/ Roskilde

1 Foo Fighters 3,269

2 Arcade Fire 2,053

3 Phlake 2,046

4 Karl William 1,738

5 Emil Stabil 1,734

6 The Hellacopters 1,719

7 Sort Sol 1,574

8 The Weeknd 1,349

9 Royal Bood 1,226

10 Anthrax 1,103

11 Justice 1,076 12 Father John Misty 1,050 13 Trentemöller 956 14 The XX 946 15 Carl Emil Petersen 919

Figure 5.7: Prediction of concert attendance at Roskilde Festival 2017 through set-based artist overlaps with Roskilde Festival Facebook page

78 Chapter 5. Evaluation page. The most popular artist according to its Roskilde Facebook audience overlap is Foo Fighters with 3,269 users. Arcade Fire and Phlake are second and third most popular artists.

Several hundred artist Facebook walls have been fetched and compared for this case study. In line with Gartner’s three types of descriptive, predictive, and prescrip-tive data analytics which were presented in the introductory chapter, this case study not only performs descriptive analytics but also shows predictive character.

Actionable insights into crowd safety operations were presented to the festival management and preparations were made for larger-than-expected crowd movement between stages. Furthermore, concert attendance was predicted and it was fore-casted that Foo Fighters and Arcade Fire will draw the largest audiences at the festival. This prediction turned out not the be correct, as The Weeknd, an artist that I ranked 8th in my prediction, drew the largest audience of the festival. A further validation of these findings is only tangentially related to the core of this thesis, and has therefore been postponed as future work.

This case study contributes to the first research question of this thesis by provid-ing an example use case for predictive analytics usprovid-ing the Social Set Visualizer. The software tool and its set-based approach have been utilized to generate meaning-ful insights in the festival analytics case, hence their predictive utility in real-world scenarios is showcased.