论文部分内容阅读
根据移动通信话务量的时间序列,采用基于模拟退火(SA)算法对超参数选择的支持向量回归机(SVR)进行建模预测。比较ARIMA、人工神经网络和SVR3种模型的预测效果,并对比研究网格法、遗传算法和SA3种SVR超参数选择方法对预测效果的影响。实验结果表明,SA-SVR预测精度高、耗时少,是一种预测移动通信话务量的有效方法。
According to the time series of mobile communication traffic, SVR (Support Vector Regression Machine) based on Simulated Annealing (SA) algorithm was chosen to model the prediction. The prediction effects of ARIMA, artificial neural network and SVR were compared. The effects of gridding, genetic algorithm and SA3 hyperspectral parameter selection on prediction were compared. The experimental results show that the SA-SVR has the advantages of high accuracy and less time consumption, which is an effective method to predict the traffic volume of mobile communication.