基于SARIMA模型的视频播放量预测软件设计与实现

发布时间：2018-01-24 10:47

本文关键词： 时间序列分析 SARIMA模型视频播放量预测　出处：《电子科技大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着计算机网络的发展与普及,以新一代网络技术、数字技术、信息技术为基础的网络视频业务,给传统媒体带来巨大的冲击,视频网站的网络流量和用户数量迅速增长,已成为互联网主要应用之一。对于视频网站而言,最主要的广告销售模式是CPM (Cost Per Mille,广告显示一千次所付费用),广告的库存和视频的播放量(video view)成线性关系。利用时间序列分析相关理论和方法,对视频播放量建立合适模型,预测未来视频播放量,从而估计未来广告库存量,以此做出正确的广告投放量判断,对网络视频门户网站有重要的现实意义。在时间序列相关理论和方法的指导下,通过对大量实际视频播放量数据的统计分析,并结合实际情况,本文选用SARIMA模型为视频播放量预测模型。结合软件设计功能需求和SARIMA模型的相关时间序列分析理论,提出了本文采用的视频播放量SARIMA模型建模及预测流程和算法实现。其中包括视频播放量数据的预处理(本文主要是异常检测及处理)、SARIMA模型识别、基于广义最小二乘法的参数估计和基于最佳预测原理的数据预测。本文选用Java语言为软件开发语言,Eclipse为Java集成开发环境,MongoDB为数据库,并在程序中调用R软件进行相关统计计算,开发一个视频播放量自动建模及预测软件。软件主要包括以下几个功能模块：数据采集模块、数据提取模块、数据转化模块、预测模块和数据存储模块。将预测值和实际值比对,分析预测效果,并通过对预测模型的诊断检验,证明采用SARIMA模型并利用本文提出的数据处理、模型识别、参数估计和预测算法能满足实际的预测要求。
[Abstract]:With the development and popularization of computer network, the network video service based on the new generation of network technology, digital technology and information technology has brought great impact to the traditional media. The rapid growth of network traffic and the number of users of video websites has become one of the main applications of the Internet. The main advertising sales model is CPM cost Per Mille.Advertising shows the fees paid for 1,000 times. The inventory of the advertisement is linearly related to the video view of the video. By using the theory and method of time series analysis, a suitable model is established to predict the amount of video playback in the future. Therefore, it is of great practical significance to estimate the amount of advertising inventory in the future and to make a correct judgment on the amount of advertising. Under the guidance of the theory and method of time series correlation, through the statistical analysis of a large number of actual video playback data, and combined with the actual situation. In this paper, SARIMA model is selected as the prediction model of video playback, combining the functional requirements of software design and the related time series analysis theory of SARIMA model. The modeling and prediction flow and algorithm of video playback SARIMA model are presented in this paper, including the preprocessing of video playback data (mainly anomaly detection and processing). SARIMA model recognition, parameter estimation based on generalized least square method and data prediction based on optimal prediction principle. This paper selects Java as the software development language and Java as the integrated development environment for the database, and calls R software in the program to carry on the related statistical calculation. The software includes the following modules: data acquisition module, data extraction module, data conversion module. Prediction module and data storage module. Compare the predicted value with the actual value, analyze the prediction effect, and through the diagnostic test of the prediction model, prove to use the SARIMA model and the data processing proposed in this paper. Model identification, parameter estimation and prediction algorithm can meet the actual prediction requirements.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP311.52

【参考文献】