基于人工神经网络的社交网站文章热度分类研究

Research on Classification of the Popularity of Social Networking Sites' Articles

社交网站作为一种新时期的交流平台,给人们的学习和生活带来了无尽的便利,逐渐成为人们获取知识,共享信息的主要渠道,但与此同时,网络文章纷繁复杂,造成用户浏览上的困扰,由此可见,对文章热度进行分类十分必要.针对这一问题,以Mashable社交网站为例,利用UCI中Online News Popularity数据集,提取文章相关属性,给出热度的评价标准.对60项属性进行了主成分分析,筛选出关键性影响因子.通过对BP神经网络和RBF神经网络两种算法进行对比研究,旨在选择一种速度更快、分类更精确的算法,结果表明,RBF神经网络的分类准确率达到94.5%,模型指标R2达到0.85,具有更好的分类表现.

In this paper,the social networking sites- Mashable is taken for an example. The Online News Popularity data- sets from UCI datasets is acquired and the relevant attributes is extracted. At the same time,the classification criteria is described. In order to analyze the correlations between the 60 features,the principal component analysis is used,and some most important features are extracted. In order to get the more accurate and faster algorithm,the two artificial neutral network separately is used and a comparison of the two algorithms is made. The experimental results indicate that RBF neutral network,whose classification accuracy rate is 94.5% and the R2 of the model is 0.85,get the better prediction performance.