Measuring Time-Constrained Influence to Predict Adoption in Online Social Networks

Recently, there has been strong interest in measuring influence in online social networks. Different measures have been proposed to predict when individuals will adopt a new behavior, given the influence produced by their friends. In this article, we show that one can achieve significant improvement over these measures, extending them to consider a pair of time constraints that provide a better proxy for social influence. By conducting an engineering study that investigates retweet networks from Twitter and Sina Weibo datasets, we tune those two parameters while we examine the correlation between influence and the probability of adoption, as well as the ability to predict adoption, estimating the real susceptibility and influence to which microblog users are dynamically subjected. Although there are limitations about using retweets to analyze social influence, our results show that for the simple count of active neighbors, its correlation with the probability of adoption is boosted up to 518.75%, whereas similar gains are observed for the other influence measures analyzed. We also obtain up to 18.89% improvement in F1 score when compared to recent machine learning techniques that aim to predict adoption, enabling practical use of the corresponding concepts for social influence applications. © 2020 ACM.