Earticle

초록

영어

A discriminative learning model with a rich feature set was used for extreme precipitation forecasting of a river basin to effectively use available historical observation data. Discrimination models use decision functions that are suitable for forecasting in cases where factors are complex or even unknown. In this study, we used the neural network technique that belongs to the empirical risk minimization method, and a support vector machine that belongs to the structural risk minimization method to be constructed on rich feature sets of rainfall in the Dongjiang basin. The results of forecast experiments show that our method was more effective compared with four types of traditional time series methods and the Naïve Bayes method that belongs to generative models. The support vector machine yielded the maximum F1 value.

목차

Abstract
 1. Introduction
 2. Data
  2.1. Basin Daily Grid Precipitation Dataset
  2.2. Basin Meteorological Data
  2.3. Typhoon Data
  2.4. Basin Meteorological Knowledge
 3. Discriminative Model
  3.1 Neural Networks
  3.2. Support Vector Machines
 4. Rich Feature Set
 5. Experiment
  5.1. Experiment Description
  5.2. Performance Measures
  5.3. Comparative Approach
  5.4. Results and Analysis
 6. Conclusion
 References

키워드

저자

• Chenming Li [ College of Computer and Information, Hohai University, Nanjing, China ]
• Xuezhen Li [ College of Computer and Information, Hohai University, Nanjing, China ]
• Jia Zhao [ School of Information Engineering, Nanchang Institute of Technology, Nanchang, China ]
• Lizhong Xu [ College of Computer and Information, Hohai University, Nanjing, China ] Corresponding author

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  International Journal of Control and Automation

• 간기

  월간

• pISSN

  2005-4297

• 수록기간

  2008~2016

• 십진분류

  KDC 505 DDC 605