论文部分内容阅读
准确预测节目流行度是互联网电视节目系统设计与优化所要解决的关键问题之一。针对现有预测方法存在模型训练时间长、样本数量多、且对突发热点节目流行度预测效果差等问题,该文测量了某互联网电视平台280万用户的60亿条收视行为数据,采用行为动力学分类方法将节目流行度演化过程分为内源临界、内源亚临界、外源临界和外源亚临界4种类型,运用双种群粒子优化的最小二乘支持向量机对每种类型分别构建了一种互联网电视节目流行度预测模型BD3P,并将BD3P模型应用于实际数据测验。实验结果表明,与现有其他方法相比,BD3P模型预测精度可提升17%以上,并能有效缩短预测周期。
Accurately forecasting program popularity is one of the key issues to be solved in the design and optimization of internet television programming systems. Aiming at the problems of long training time, large number of samples and poor forecasting popularity of sudden hot spots in the existing prediction methods, this paper measured 6 billion pieces of viewing behavior data of 2.8 million users on an Internet television platform, Kinetic classification method divides program evolution into four types: endogenous criticality, endogenous subcritical, exogenous criticality and exogenous subcritical. Two-species particle-optimized least squares support vector machine This paper constructs a BD3P forecasting model of Internet TV program popularity, and applies the BD3P model to the actual data test. The experimental results show that compared with other existing methods, the prediction accuracy of BD3P model can be improved by more than 17%, and the prediction period can be shortened effectively.