Limits...
Predictability of extreme events in social media.

Miotto JM, Altmann EG - PLoS ONE (2014)

Bottom Line: It is part of our daily social-media experience that seemingly ordinary items (videos, news, publications, etc.) unexpectedly gain an enormous amount of attention.Here we investigate how unexpected these extreme events are.This indicates that, despite the inherently stochastic collective dynamics of users, efficient prediction is possible for the most successful items.

View Article: PubMed Central - PubMed

Affiliation: Max Planck Institute for the Physics of Complex Systems, Dresden, Germany.

ABSTRACT
It is part of our daily social-media experience that seemingly ordinary items (videos, news, publications, etc.) unexpectedly gain an enormous amount of attention. Here we investigate how unexpected these extreme events are. We propose a method that, given some information on the items, quantifies the predictability of events, i.e., the potential of identifying in advance the most successful items. Applying this method to different data, ranging from views in YouTube videos to posts in Usenet discussion groups, we invariantly find that the predictability increases for the most extreme events. This indicates that, despite the inherently stochastic collective dynamics of users, efficient prediction is possible for the most successful items.

Show MeSH
Predictability increases for extreme events.If the attention an item receives at time  is above a threshold, , an event  is triggered. The plots show how the predictability  changes with  using two different informations to combine the items in groups . Black circles:  at time  using metadata of the items to group them. The red lines are computed using as probabilities  the Extreme Value distribution fits for each group at a threshold value , see Eq. (1) and SI Sec. 2. Blue squares:  at time  using , i.e., the attention the item obtained at day . The dashed lines are the values of the 95% percentile of the distribution generated by measuring  in an ensemble of databases obtained shuffling the attribution of groups () to items (the colors match the symbols and symbols are shown only where  is at least twice this value). Results for the four databases are shown: (a) YouTube (: views of a video; metadata: video category); (b) Usenet discussion groups (: posts in a thread; metadata: discussion group of the thread); (c) Stack-Overflow (: votes to a question; metadata: programming language of the question, see SI Sec. 2 for details); (d) PLOS ONE (: online views of a paper; metadata: number of authors of the paper).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4219754&req=5

pone-0111506-g003: Predictability increases for extreme events.If the attention an item receives at time is above a threshold, , an event is triggered. The plots show how the predictability changes with using two different informations to combine the items in groups . Black circles: at time using metadata of the items to group them. The red lines are computed using as probabilities the Extreme Value distribution fits for each group at a threshold value , see Eq. (1) and SI Sec. 2. Blue squares: at time using , i.e., the attention the item obtained at day . The dashed lines are the values of the 95% percentile of the distribution generated by measuring in an ensemble of databases obtained shuffling the attribution of groups () to items (the colors match the symbols and symbols are shown only where is at least twice this value). Results for the four databases are shown: (a) YouTube (: views of a video; metadata: video category); (b) Usenet discussion groups (: posts in a thread; metadata: discussion group of the thread); (c) Stack-Overflow (: votes to a question; metadata: programming language of the question, see SI Sec. 2 for details); (d) PLOS ONE (: online views of a paper; metadata: number of authors of the paper).

Mentions: The examples above show that formula (2) allows for a quantification of the importance of different factors (e.g., number of authors, early views to the paper) to the occurrence of extreme events, beyond correlation and regression methods (see also Ref. [19]). Besides the quantification of the predictability of specific problems, by systematically varying and we can quantify how the predictability changes with time and with event magnitude. Our most significant finding is that in all tested databases and grouping strategies the predictability increases with , i.e., extreme events become increasingly more predictable, as shown in Fig. 3.


Predictability of extreme events in social media.

Miotto JM, Altmann EG - PLoS ONE (2014)

Predictability increases for extreme events.If the attention an item receives at time  is above a threshold, , an event  is triggered. The plots show how the predictability  changes with  using two different informations to combine the items in groups . Black circles:  at time  using metadata of the items to group them. The red lines are computed using as probabilities  the Extreme Value distribution fits for each group at a threshold value , see Eq. (1) and SI Sec. 2. Blue squares:  at time  using , i.e., the attention the item obtained at day . The dashed lines are the values of the 95% percentile of the distribution generated by measuring  in an ensemble of databases obtained shuffling the attribution of groups () to items (the colors match the symbols and symbols are shown only where  is at least twice this value). Results for the four databases are shown: (a) YouTube (: views of a video; metadata: video category); (b) Usenet discussion groups (: posts in a thread; metadata: discussion group of the thread); (c) Stack-Overflow (: votes to a question; metadata: programming language of the question, see SI Sec. 2 for details); (d) PLOS ONE (: online views of a paper; metadata: number of authors of the paper).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4219754&req=5

pone-0111506-g003: Predictability increases for extreme events.If the attention an item receives at time is above a threshold, , an event is triggered. The plots show how the predictability changes with using two different informations to combine the items in groups . Black circles: at time using metadata of the items to group them. The red lines are computed using as probabilities the Extreme Value distribution fits for each group at a threshold value , see Eq. (1) and SI Sec. 2. Blue squares: at time using , i.e., the attention the item obtained at day . The dashed lines are the values of the 95% percentile of the distribution generated by measuring in an ensemble of databases obtained shuffling the attribution of groups () to items (the colors match the symbols and symbols are shown only where is at least twice this value). Results for the four databases are shown: (a) YouTube (: views of a video; metadata: video category); (b) Usenet discussion groups (: posts in a thread; metadata: discussion group of the thread); (c) Stack-Overflow (: votes to a question; metadata: programming language of the question, see SI Sec. 2 for details); (d) PLOS ONE (: online views of a paper; metadata: number of authors of the paper).
Mentions: The examples above show that formula (2) allows for a quantification of the importance of different factors (e.g., number of authors, early views to the paper) to the occurrence of extreme events, beyond correlation and regression methods (see also Ref. [19]). Besides the quantification of the predictability of specific problems, by systematically varying and we can quantify how the predictability changes with time and with event magnitude. Our most significant finding is that in all tested databases and grouping strategies the predictability increases with , i.e., extreme events become increasingly more predictable, as shown in Fig. 3.

Bottom Line: It is part of our daily social-media experience that seemingly ordinary items (videos, news, publications, etc.) unexpectedly gain an enormous amount of attention.Here we investigate how unexpected these extreme events are.This indicates that, despite the inherently stochastic collective dynamics of users, efficient prediction is possible for the most successful items.

View Article: PubMed Central - PubMed

Affiliation: Max Planck Institute for the Physics of Complex Systems, Dresden, Germany.

ABSTRACT
It is part of our daily social-media experience that seemingly ordinary items (videos, news, publications, etc.) unexpectedly gain an enormous amount of attention. Here we investigate how unexpected these extreme events are. We propose a method that, given some information on the items, quantifies the predictability of events, i.e., the potential of identifying in advance the most successful items. Applying this method to different data, ranging from views in YouTube videos to posts in Usenet discussion groups, we invariantly find that the predictability increases for the most extreme events. This indicates that, despite the inherently stochastic collective dynamics of users, efficient prediction is possible for the most successful items.

Show MeSH