Learning to predict trending queries

摘要: Among the many tasks driven by very large scaled web search queries, it is an interesting task to predict how likely queries about a topic become popular (a.k.a. trending or buzzing) as news in near future, which known "Detecting queries." This nontrivial since realization of buzzing trends often requires sufficient statistics through users' activities. To address this challenge, we propose novel framework that predicts whether future. In principle, our system built on two learners. The first learn dynamics time series for queries. second, decision maker, binary classifier determines trending. Our extremely efficient be taking advantage grid architecture allows deal with volume data. addition, flexible continuously adapt patterns evolve. experiments results show approach achieves high quality accuracy (over 77.5%} true positive rate) and yet detects much earlier (on average 29 hours advanced) than baseline system.