Social networking websites allow users to create and share content. Biginformation cascades of post resharing can form as users of these sites reshareothers' posts with their friends and followers. One of the central challengesin understanding such cascading behaviors is in forecasting informationoutbreaks, where a single post becomes widely popular by being reshared by manyusers. In this paper, we focus on predicting the final number of reshares of agiven post. We build on the theory of self-exciting point processes to developa statistical model that allows us to make accurate predictions. Our modelrequires no training or expensive feature engineering. It results in a simpleand efficiently computable formula that allows us to answer questions, inreal-time, such as: Given a post's resharing history so far, what is ourcurrent estimate of its final number of reshares? Is the post resharing cascadepast the initial stage of explosive growth? And, which posts will be the mostreshared in the future? We validate our model using one month of completeTwitter data and demonstrate a strong improvement in predictive accuracy overexisting approaches. Our model gives only 15% relative error in predictingfinal size of an average information cascade after observing it for just onehour.
展开▼
