基于改进人口模型的微博话题趋势预测  

Based on improved malthusian model microblogging topic trend forecast

作　　者：何炎祥[1,2] 刘健博[1,2] 刘楠[3] 彭敏[1,2] 陈强[1,2] 何静 

机构地区：[1]武汉大学计算机学院,湖北武汉430072 [2]武汉大学软件工程国家重点实验室,湖北武汉430072 [3]军事经济学院军需系,湖北武汉430035 [4]肯尼索州立大学,美国30144

出　　处：《通信学报》2015年第4期1-8,共8页Journal on Communications

基　　金：国家自然科学基金资助项目(61472290;61472291);国家自然科学基金青年基金资助项目(61303115);武汉市科技攻关基金资助项目(201210421135)~~

摘　　要：趋势预测问题是当前社会计算中的研究热点,但微博平台的话题趋势预测还处在探索阶段。在分析、抽取和定义事件趋势的影响因素的基础上,改进了人口模型使其适用于微博趋势预测,然后将改进的人口模型映射在神经网络上,并利用遗传优化的神经网络对事件的趋势进行预测。实验证明,该方法对于预测网络中长期酝酿的事件发展趋势效果明显,能够有效地预测事件的爆发点和发帖量,而且适用于小样本预测问题。Although trend forecasting problem is the research focus of social computing, yet the topic trend forecast of micro logging platform is still at the exploratory stage. Based on the analyzing, extracting and defining the influencing factors of the trends of event, the improved Malthus model, which is more applicable to solve microblogging topic trend forecast problem is presented. Then the model is mapped to the neural network, and the genetic algorithms are used to predict the trend of events. Experiments show that this method can predict the trend of long-term events in the social net- work with good results, and can predict the outbreak point and posting volume effectively. It is worth nothing that the method is also suitable for small sample forecasting problems.

关 键 词：社会计算 趋势预测:神经网络 人口模型 社交网络 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]