基于小波分析的保险收入预测  

Forecast of insurance income based on FFT and wavelet analysis

作　　者：刘彬生[1] 李一军[1] 

机构地区：[1]哈尔滨工业大学管理学院,哈尔滨 150001

出　　处：《哈尔滨工程大学学报》2006年第B07期152-157,共6页Journal of Harbin Engineering University

基　　金：国家自然科学基金资助项目（70471027）.

摘　　要：提出了一种基于小波分析组合预测模型：首先利用Mallat算法对保险收入时间序列进行多尺度分解，得到对应尺度下的概貌（低频）分量和细节（高频）分量，然后从分解中提取趋势项并建模，最后对周期项（含季节项）和随机项进行了建模探讨，指出由于部分小波分解项所含有的周期存在相关性，此时对每一项分开建模并不一定能提高预测精度，通过给出分解项合并原则，然后对部分分解项进行合并建模．最后将各建立模型的预测结果进行叠加即可得原保险收入变量的预测值．将该模型用于中国保险收入的预测中，并与传统预测模型ARIMA进行了对比分析，结果表明，建立的组合模型充分利用了现有信息，预测精度高．A combination forecast model based on the Fast Fourier Transform （FFT） and wavelet analysis is proposed. First, the time series of monthly economics data is decomposed under a multi-scale by using the Mallat algorithm. So the approximate （low frequency） component and the detailed （high frequency） component under the corresponding scale are obtained. Then, the trend is extracted from the decomposition and modeled separately. Through the discussion about the cycle and stochastic item, it can be seen that modeling each decomposition item separately can＇t lead to more precise forecast result because of the cycle correlation between the decomposition items. So a method is given on how to combine the decomposition items based on FFT. Then each combination item is modeled. Finally, the superimposition of forecasts result can forecast the original insurance income. By contrast between the forecast result and the traditional forecast model ARIMA with the monthly data of Chinese insurance income, it indicates that this model makes full use of the existing information and is more precise in forecast.

关 键 词：小波分析 MALLAT算法 快速傅立叶变换 增长曲线 

分 类 号：F224.0[经济管理—国民经济]