论文部分内容阅读
针对同一地区不同时间内直播星电视服务用户的数量变化特征,该文提出了基于时间序列的直播星用户预测方法,并通过ARIMA模型预测用户数量。分析了ARIMA模型的建立方法和评价参数,对时间序列模型ARIMA(p,d,q)的参数进行不同组合的尝试;利用模型评价参数(平稳的R方和正态化的BIC)确定模型的阶数p、q;对模型进行参数检验和显著性检验,根据检验结果确立最终模型为ARIMA(4,2,3);采用宁夏用户数据对模型拟合效果进行验证,进而对未来几个月的用户数量进行预测。结果表明,时间序列模型ARIMA(4,2,3)对宁夏用户数量变化的预测准确度比较高。
Aiming at the characteristics of the number of live satellite TV service users in different time in the same area, this paper proposes a time series based direct satellite user prediction method and predicts the number of users by ARIMA model. The methods of ARIMA model establishment and evaluation parameters are analyzed, and different combinations of parameters of ARIMA (p, d, q) model are attempted. The model evaluation parameters (stationary R-square and normalized BIC) Order p, q; parameter test and significance test of the model, according to the test results to establish the final model ARIMA (4,2,3); Ningxia user data to verify the model fitting effect, and then for the next few months Of the number of users to predict. The results show that ARIMA (4, 2, 3) model of time series has relatively high accuracy in predicting the number of users in Ningxia.