Research on Temperature Forecasts Based on Wavelet Neural Network

Research on pricing of weather derivatives is of great importance for enhancing agricultural and industrial safety. Temperature index is viewed as the core of the pricing of weather derivatives. This paper is prepared mainly with reference to the inherent characteristics that may be found in the temperature signals. The Wavelet Neural Network (WNN) is characterized by a combination of wavelet multi-resolution analysis with neural network nonlinear approximation. By application of the temperature data of Chongqing in the period from 1951 to 2010, the wavelet neural network model was used for temperature prediction and estimates. The empirical results of simulation and verification showed that the wavelet neural network established in this paper was capable of performing better prediction of future temperatures, and thus could be applied to areas such as pricing of weather derivatives. (C) 2011 Published by American Applied Sciences Research Institute