基于随机森林的微博互动特征分析

Analysis of Interactive Characteristics of Weibo Based on Random Forest

微博凭借其开放性、低门槛已成为最常用的社交媒体平台之一,其海量数据背后蕴藏着巨大的价值亟待研究。而准确地判断微博的传播趋势,降低不良微博带来的影响已成为当前面临的主要问题。文中以新浪微博为研究对象,将随机森林算法与数据分析处理相结合,对微博的博文发布一周后的转评赞行为进行预测,将数据特征分为三类并分析了每类特征对预测结果的影响。首先,简述了决策树及随机森林算法的原理;其次,对微博数据进行分析,将提取的特征分为用户特征、时间特征和文本类特征三类;最后,通过三组对比实验验证了随机森林算法在微博互动预测上的可行性,并分析了三类特征对预测结果的影响。实验结果表明,用户特征对预测准确率的影响较大。

Weibo has become one of the most commonly used social media platforms due to its openness and low threshold,and the huge value behind its massive data needs to be studied.To accurately judge the spread trend of Weibo and reduce the impact of bad Weibo has become the main problem.Taking Sina Weibo as the research object,we combine random forest algorithm with data analysis and processing to predict the behavior of the review and praise of Weibo after one week of blog post release.We divide data features into three categories and analyze the influence of each type of features on the predicted results.Firstly,the principle of decision tree and random forest algorithm is briefly described.Secondly,the microblog data is analyzed,and the extracted features are divided into three categories:user feature,time feature and text class feature.Finally,three sets of contrast experiments are verified.The feasibility of the random forest algorithm in the interactive prediction of Weibo,and the influence of the three types of features on the prediction results are analyzed.The experiment shows that the user feature has a greater impact on the accuracy of prediction.