基于小波-NAR神经网络的气象要素时间序列预测与天气指数彩虹期权估值

本文基于小波-NAR神经网络技术,提出气象要素时间序列预测与天气指数彩虹期权估值的原理与方法,同时采用2000-2014年悉尼日均气温和日降雨量数据,进行气象预测与天气期权估值.结果显示:小波-NAR神经网络因灵活的非线性动态结构较好地反映了气象变化特征,其预测与估值效果优于其他模型;该天气期权价值形成中的非线性特征取决于五种经济效应.科学预测天气和估计天气期权价值,开发天气衍生品,可挖掘天气不确定性的经济价值,弱化其对天气敏感产业的影响.

Forecasting of meteorological time series and pricing of weather index rainbow options: A wavelet-NAR neural network model

Based on the technique of wavelet-NAR neural network, this paper develops a method for forecasting meteorological elements and pricing weather index rainbow options. By using a data set of daily average temperature and rainfall in Sydney from 2000 to 2014, an empirical analysis of the meteorological forecast and weather option pricing is conducted. The results show that the wavelet-NAR neural network model is more accurate for forecasting meteorological time series and pricing weather index rainbow options because the flexible nonlinear dynamic structure of the model can better reflect the meteorological characteristics. This study also finds that the nonlinearity in the formation of the weather option value is determined by five economic effects. Our findings suggest that a scientific weather forecast and weather option pricing as well as a development of weather derivatives can contribute to mining the economic value of weather uncertainty and weakening its impact on weather-sensitive industries.