短时间序列集的一种预测调和方法

A forecasting adjustment method for the set of short time series

由于互联网以及大数据产业的高速发展,各行业产生了大量的短时间序列数据。因此,对这些数据进行分析进而预测其未来趋势成为了重要的生产和管理的手段。短时间序列以单个序列的观测数量少为特征,是时间序列分析的一个难点。如果预测对象是短时间序列数据集,就可以利用其总量的预测值去调节各分量的预测值。文章提出了一种时间序列的预测调和方法,并通过此方法去调节ARIMA模型对一个短时间序列数据集的建模预测结果,与ARIMA的预测结果相比,调和后的预测精度得到了提高。

Along with the flourishing development of the industry of the Internet and Big data, the short time series data appears in various fields. Therefore, analyzing these short time series data and forecasting its future trend is cru- cial for our production and management. The short time series characterized by small amount data of single time se- ries is difficult for time series analysis. However, the forecasting series is a set of short time series, and we can adjust the forecast results of single series through the forecast results of total amount data. In the paper, a forecasting ad- justment method for time series is proposed with using the method to adjust the forecasting results through modeling a set of short time series by ARIMA model. The forecast accuracy has been adjusted for improving ARIMA model.