中文摘要：

交互式社会媒体上的热点话题具有巨大的影响力,对热点话题进行建模和预测是一个非常重要但困难的问题.针对话题参与用户的特点进行了分析,构建了用户活跃度以及用户重入概率等模型的合理假设条件.根据话题发展模式和基于用户参与话题概率构建了单峰模型和多峰模型.分别基于两个不同数据集对模型进行了拟合和预测试验,试验结果表明,本文提出的模型在拟合与预测话题的发展趋势上的效果都优于SpikeM模型,尤其是对具有复杂波动发展模式的话题,提出的模型能很好地拟合与预测话题的波动.

英文摘要：

Hot topics on interactive social media websites enormously affect the incidence and development of the various events in both virtual and real world.Modeling and predicting information propagation process of hot topics are very important but difficult research problems.In this paper,characteristics of participants in hot topics are deeply analyzed.As a result,user activity degree,user popularity degree and user re-entrance probability are defined.The assumptions of traditional information propagation models of hot topics are relaxed according to two features in a hot topic：one user could participate the same topic many times and different users have different activity degrees.According two types of propagation patterns of hot topics,two effective models are proposed based on user participation probability.The first modelis used to model the single peak propagation pattern topics and the second mode is used to model multi-peaks propagation pattern topics.Two datasets are selected from popular social media websites and comprehensive experiments are conducted.Two models proposed in this paper and SpikeM model are implemented for comparative study.The experimental results show that the models proposed in this paper can effectively simulate single peak propagation pattern and multi-peaks propagation pattern of hot topics.Especially,the models proposed in this paper outperform SpikeM model for fitting topics with complex rise-fall propagation patterns.Furthermore,the models can accurately predict future propagation patterns of hot topics in real datasets.