一种基于视频特征及历史数据的流行度预测算法

摘要

针对流媒体的流行度预测问题,提出一种基于视频特征及历史数据的流行度预测模型.首先,根据视频特征及在社交网络中的影响力,使用K-近邻(KNN)算法对视频的流行程度进行预测.然后,基于流行程度的预测结果,结合自回归滑动平均(Autoregressive Moving Average,ARMA)模型对视频的点播量进行预测.最后,通过爬取豆瓣电影及新浪微博数据,对模型进行试验.结果表明,与朴素贝叶斯分类器及ARMA模型相比,本文模型的召回率(recall)明显较高,平均平方根误差(RMSE)降低了约20％.%For the popularity prediction of streaming media,a model of popularity prediction based on video characteristics and historical data is proposed.Firstly,according to the video characteristics and the influence in the social network,the popularity of video is predicted using K-Nearest Neighbor(KNN).Subsequently,based on the results of last step,combined with the Autoregressive Moving Average (ARMA) model,the on-demand quantity of the video is predicted using historical data.Finally,the experiment is carried out by crawling the Douban film and Sina microblogging data.The results show that the recall rate of the model is higher than that of the Naive Bayesian classifier,and the average square root error (RMSE) is decreased by about 20％,compared with the ARMA model.