基于集成学习的微博用户转发行为预测

为了提高微博用户转发行为预测的精度,提出一种有效的基于集成学习的微博用户转发行为预测算法.首先,对影响用户转发的各种特征进行综合分析,提取出用户属性、社交关系、微博内容等影响用户转发行为的特征;然后,采用Logistic回归、支持向量机与BP(BackPropagation)神经网络等机器学习算法对用户转发行为进行预测;最后,利用"加权投票法"的集成学习方法对多个预测结果进行融合.实验结果表明,相对于BP神经网络算法,在综合评价性能的F1度量值上,集成学习算法有1.5%的性能提升.

Predicting microblog user retweet behaviors based on ensemble learning

In order to improve the accuracy of predicting userretweet behaviors in a micorblog social network,the paper proposes an effective method based on Ensemble Learning.Firstly,the papercomprehensively analyzes the performances of various features that affect user retweetbehaviors,such asuser attributes,social relationships and microblog contents,et al.Based on the extracted features,the propsed method respectively predictsuser retweetbehaviors with Logistic regression,SVM(Support Vector Machine)and BP(Back Propagation)neural network,and incorporates the corresponding results in a weighted voting manner based on Ensemble Learning.The experimental results show that the propsed method has a performance improvement of 1.5% on F1 metric of the overall evaluation,compared with the BP neural network.