作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2011, Vol. 37 ›› Issue (2): 178-179. doi: 10.3969/j.issn.1000-3428.2011.02.061

• 人工智能及识别技术 • 上一篇    下一篇

基于SVR与微分进化策略的话务量预测

韩 锐1,贾振红1,覃锡忠1,常 春2,王 浩2   

  1. (1. 新疆大学信息科学与工程学院,乌鲁木齐 830046;2. 中国移动新疆分公司,乌鲁木齐 830063)
  • 出版日期:2011-01-20 发布日期:2011-01-25
  • 作者简介:韩 锐(1984-),男,硕士研究生,主研方向:人工智能,模式识别;贾振红,教授、博士生导师;覃锡忠,副教授;常 春,高级工程师、硕士;王 浩,工程师、硕士
  • 基金资助:

    中国移动新疆分公司研究发展基金资助项目

Telephone Traffic Load Prediction Based on SVR with DE-strategy

HAN Rui 1, JIA Zhen-hong 1, QIN Xi-zhong 1, CHANG Chun 2, WANG Hao 2   

  1. (1. College of Information Science and Engineering, Xinjiang University, Urumqi 830046, China; 2. Xinjiang Mobile Communication Company, Urumqi 830063, China)
  • Online:2011-01-20 Published:2011-01-25

摘要:

采用支持向量回归机(SVR)与微分进化策略相结合的方法,对新疆2个地区的月平均忙时话务量进行预测。由微分进化策略良好的全局搜索性质,以预测平均相对误差为目标函数,对SVR的超参数进行寻优,利用优化后的SVR月平均忙时话务量进行预测。与传统的网格寻优算法和RBF神经网络方法进行比较,结果表明,SVR的泛化能力与微分进化策略的搜索能力相结合,可以得到更好的预测 效果。

关键词: 微分进化策略, 支持向量回归机, 话务量预测

Abstract:

Telephone traffic load of monthly busy hour in two states of Xinjiang are predicted by the method of Support Vector Regression(SVR) combining with Differential Evolution strategy(DE-strategy). The hyper-parameter of SVR is optimized via the DE-strategy and the MAPE criteria is defined as the objective function. Telephone traffic load of monthly busy hour is forecasted by the optimized SVR, the predicted result is compared with the method of grid search and RBF neural network. A better prediction result is obtained by the generalization property of SVR combining with searching property of DE-strategy.

Key words: Differential Evolution strategy(DE-strategy), Support Vector Regression(SVR), telephone traffic load prediction

中图分类号: