摘要
为了提高组合预测的精度,提出了一种新的组合权重计算方法,该方法通过将平均绝对百分数误差(MAPE)和最小二乘法相结合来确定组合预测模型的权重值。将这种新的组合权重方法应用到组合模型中,并对湖北省国内生产总值(GDP)进行预测。首先,建立了差分自回归移动平均(ARIMA)模型和指数曲线回归模型;然后,用MAPE和最小二乘法确定组合模型的权系数,在此基础上将两种权系数进行组合,形成组合权重。预测结果表明:该组合权重与单一权重相比,可将组合模型的预测精度提高约0.3%。
In order to improve the prediction accuracy,a new method to calculate the combination weights was proposed. The weight value of the combination model was determined by combining the mean absolute percentage error( MAPE) with the least square method.The new combination weights method was applied to a combination model,and the gross domestic product( GDP) of Hubei province was forecasted by using the combination model. Firstly,the autoregressive integrated moving average( ARIMA) model and exponential curve regression model were established. Then MAPE and least square method were used to determine the weight coefficient of the combined model.On this basis,the two weight coefficients were combined to form the combined weight. The predication results show that the prediction accuracy of the combined model with the combined weight is improved about 0.3% compared with the single weight.
出处
《河南科技大学学报（自然科学版）》
CAS
北大核心
2018年第2期87-93,98,共8页
Journal of Henan University of Science And Technology:Natural Science
基金
国家自然科学基金项目(91324201)
