计算机应用 ›› 2017, Vol. 37 ›› Issue (9): 2689-2693.DOI: 10.11772/j.issn.1001-9081.2017.09.2689

• 应用前沿、交叉与综合 • 上一篇    下一篇

基于改进的Adaboost-BP模型在降水中的预测

王军1,2, 费凯1, 程勇2   

  1. 1. 南京信息工程大学 计算机与软件学院, 南京 210044;
    2. 南京信息工程大学 信息化建设与管理处, 南京 210044
  • 收稿日期:2017-03-17 修回日期:2017-05-18 出版日期:2017-09-10 发布日期:2017-09-13
  • 通讯作者: 费凯,feikainiust@163.com
  • 作者简介:王军(1970-),男,安徽铜陵人,教授,硕士,CCF会员,主要研究方向:无线传感网、大数据;费凯(1990-),男,江苏连云港人,硕士研究生,主要研究方向:大数据气象领域应用;程勇(1980-),男,重庆人,高级工程师,博士,CCF会员,主要研究方向:无线传感网、大数据。
  • 基金资助:
    国家自然科学基金资助项目(61402236, 61373064);江苏省农业气象重点实验室开放基金资助项目(KYQ1309);江苏省"六大人才高峰"项目(2015-DZXX-015,2013-DZXX-019);江苏省产学研前瞻性联合研究项目(BY2014007-2);公益性行业(气象)科研专项(GYHY201106037)。

Prediction of rainfall based on improved Adaboost-BP model

WANG Jun1,2, FEI Kai1, CHENG Yong2   

  1. 1. College of Computer and Software, Nanjing University of Information Science and Technology, Nanjing Jiangsu 210044, China;
    2. Information Construction and Management Department, Nanjing University of Information Science and Technology, Nanjing Jiangsu 210044, China
  • Received:2017-03-17 Revised:2017-05-18 Online:2017-09-10 Published:2017-09-13
  • Supported by:
    This work is partially supported by the National Natural Science Foundation of China (61402236, 61373064), Opening Foundation of Jiangsu Key Laboratory of Agricultural Meteorology (KYQ1309), Six Talent Peaks Project in Jiangsu Province (2015-DZXX-015, 2013-DZXX-019), the Prospective Joint Research Project of Jiangsu Province (BY2014007-2), Special Fund for Meteorological Research in Public Interest (GYHY201106037).

摘要: 针对目前分类算法对降水预测过程存在着泛化能力低、精度不足的问题,提出改进Adaboost算法集成反向传播(BP)神经网络组合分类模型。该模型通过构造多个神经网络弱分类器,赋予弱分类器权值,将其线性组合为强分类器。改进后的Adaboost算法以最优化归一化因子为目标,在提升过程中调整样本权值更新策略,以此达到最小化归一化因子的目的,从而确保增加弱分类器个数的同时降低误差上界估计,通过最终集成的强分类器来提高模型的泛化能力和分类精度。选取江苏境内6个站点的逐日气象资料作为实验数据,建立7个降水等级的预报模型,从对降雨量有影响的众多因素中,选取12个与降水相关性较大的属性作为预报因子。通过多次实验统计,结果表明基于改进的Adaboost-BP组合模型具有较好的性能,尤其对58259站点的适应性较好,总体分类精度达到81%,在7个等级中,对0级降雨的预测精度最好,对其他等级的降雨预测有不同程度的精度提升,理论推导及实验结果证明该种改进可以提高预测精度。

关键词: 分类器, 改进Adaboost, BP神经网络, 组合模型, 权值调整, 归一化因子

Abstract: Aiming at the problem that the current classification algorithm has low generalization ability and insufficient precision, a combination classification model combining Adaboost algorithm and Back-Propagation (BP) neural network was proposed. Multiple neural network weak classifiers were constructed and weighted, which were linearly combined into a strong classifier. The improved Adaboost algorithm aimed to optimize the normalization factor. The sample weight update strategy was adjusted during the lifting process, to minimize the normalization factor, increasing the number of weak classifiers while reducing the error upper bound estimate was ensured, and the generalization ability and classification accuracy of the final integrated strong classifier was improved. A daily precipitation model of 6 sites in Jiangsu province was selected as the experimental data, and 7 precipitation models were established. Among the many factors influencing the rainfall, 12 attributes with large correlation with precipitation were selected as the forecasting factors. The results show that the improved Adaboost-BP combination model has better performance, especially for the site 58259, and the overall classification accuracy is 81%. Among the 7 grades, the prediction accuracy of class-0 rainfall is the best, and the accuracy of other types of rainfall forecast is improved. The theoretical derivation and experimental results show that the improvement can improve the prediction accuracy.

Key words: classifier, improved Adaboost, BP neural network, combined model, weight adjustment, normalization parameter

中图分类号: