国际油价预测之新思想——分时间段类型研究

A New Method on International Oil Price Forecasting

站在全新的角度——在不同类型的时段分别研究油价预测模型的优劣.通过小波变换对油价序列进行奇异性分析,将油价序列所处的时间段分为两类:含奇异点时段,不含奇异点时段.并在不同类型的时段分类比较多元回归模型、延迟因变量回归模型、神经网络模型的拟合预测效果,最终得出无论何时进行油价预测,在上述模型中效果最为理想的是神经网络模型.

A new method that the comparison of pros and cons on variety models was carried out to predict the oil price in different periods.The singularity analysis on the sequences of oil＇s price was complied by wavelet transform in this article and the time period for the sequences of oil＇s price was divided into two sections： the time period with singular points and the time period with non-singular points.The multiple regressive model,delayed dependent variable regressive model and neural network model were compared about the forecasting effects and fitting effects in the two different time periods.Finally,it comes to the conclusion that the most reasonable model is neural network model at all time.