Computer Engineering and Applications ›› 2014, Vol. 50 ›› Issue (20): 211-214.

Previous Articles     Next Articles

Busy traffic forecasting based on improved semi supervised SVR algorithm

LAN Jiao1, QIN Xizhong1, JIA Zhenhong1, CHEN Li2   

  1. 1.Institute of Information Science and Engineering, Xinjiang University, Urumqi 830046, China
    2.Subsidiary Company of China Mobile in Xinjiang, Urumqi 830046, China
  • Online:2014-10-15 Published:2014-10-28

基于改进半监督SVR算法的忙时话务量预测

兰  娇1,覃锡忠1,贾振红1,陈  丽2   

  1. 1.新疆大学 信息科学与工程学院,乌鲁木齐 830046
    2.中国移动通信集团新疆有限公司,乌鲁木齐 830046

Abstract: In order to improve the operator holidays busy traffic prediction accuracy, through analyzing the holidays busy traffic data characteristics, it puts forward the improved semi supervised SVR prediction algorithm. The method uses the semi supervised learning based on graph Laplacian to deform and train SVR kernel matrix. In allusion to graph Laplacian large amount of calculation, the Nystrom algorithm is proposed to optimize it. The simulation results show that the proposed algorithm has a good generalization ability and high prediction precision.

Key words: holidays busy traffic forecasting, Support Vector Regression(SVR) machine, semi supervised learning, graph Laplacian;Nystrom algorithm

摘要: 为了提高运营商节假日忙时话务量的预测精度,通过分析各节假日忙时话务量数据的特点,提出基于改进半监督支持向量机预测算法。该方法采用基于图形拉普拉斯算子的半监督学习算法来变形训练支持向量回归机的核矩阵。针对图形拉普拉斯算子计算量较大的问题,采用Nystrom算法对其进行优化。仿真结果表明,提出的算法有较好的泛化能力和较高的预测精度。

关键词: 节假日忙时话务预测, 支持向量回归机, 半监督学习, 图形拉普拉斯算子, Nystrom算法