Facebook Cascade Prediction
In class we have discussed information cascades and how they begin. In an information cascade, a person makes a decision based off of the observation of other people’s decisions. In this situation, if enough people make the same decision, everyone around them will make the same decision and therefore there will be an information cascade.
Shares on Facebook are just one example of information cascades in everyday life. As people share photos amongst there friends, cascades of reshares can form. The Facebook Data Science Team looked into whether or not one can predict a cascade. They wanted to answer the question of “what’s the difference between a photo that is shared hundreds, or even tens of thousands of times, and one that’s shared only a few times?”. To answer this question, Facebook traked the growth and cascades of photo reshares on their website. Some of the factors they considered to predict whether the cascade will grow large were: “content features (e.g. whether a photo was taken indoors, or contained food), user features (e.g. number of friends), structural features (e.g. how deep a cascade tree grew), and temporal features (e.g. how fast a photo spread”. Facebook then created a machine learning algorithm using these features to predict whether the cascade will double in size. This algorithm had 80% accuracy. Interestingly, as the cascade gets larger, the features that are most important to the prediction change. For example, when the photo has more reshares, information about the user who posted the photo matters less, while information about how fast the photo is being reshared becomes more relevant.
The two most important factors in this prediction were the speed at which the cascade had been shared in the past and how “deep” the cascade is. A cascade being deep refers to how many people outside immediate friends the photo spreads too. For example, if at least one of the photos first few shares was by a friend of a friend then the photo has a higher probability of spreading further. If a photo spreads further away from you, this is suggestive that the content is interesting to other and is not specific to just you and your friends. Facebook is using this information about the spread of information cascades to gain a greater understanding of information spread on the internet and also to help identify trendy content in its early stages.
https://www.facebook.com/notes/facebook-data-science/can-cascades-be-predicted/10152056491448859/