论文部分内容阅读
目的对基于时间序列的三种预测模型即自回归滑动平均混合模型(ARIMA)、灰色模型(GM)、广义回归神经网络模型(GRNN)进行尘肺发病预测的适用性比较。方法选用河北省1954—2015年62年的尘肺发病数据,前54年数据用来拟合预测,后8年数据来比较三种模型的预测效果;采用预测误差(prediction error,PE)、平均绝对误差(mean absolute error,MAE)和平均相对误差(mean relative error,MRE)评价拟合效果。结果 GM(1,1)的预测结果较差,ARIMA的MAE和MRE是三种模型中最小的,其短期预测的PE也最低;三种方法长期预测的PE都比较大,比较而言GRNN的长期预测结果最好。结论 ARIMA适用于尘肺发病的短期预测,GRNN适用于长期预测。
OBJECTIVE To compare the applicability of three prediction models based on time series (ARIMA), gray model (GM) and generalized regression neural network model (GRNN) to predict the incidence of pneumoconiosis. Methods The pneumoconiosis data of 62 years from 1954 to 2015 in Hebei Province were selected. The data of the first 54 years were used to fit the prediction and the data of the later 8 years to compare the prediction results of the three models. The prediction error (prediction error, PE) The mean absolute error (MAE) and mean relative error (MRE) were used to evaluate the fitting effect. Results The prediction results of GM (1,1) were poor. The MAE and MRE of ARIMA were the smallest among the three models, and the PE of short-term prediction was also the lowest. The long-term prediction of PE by these three methods was relatively large. In comparison, GRNN Long-term forecast results best. Conclusion ARIMA is suitable for the short-term prediction of pneumoconiosis and GRNN is suitable for long-term prediction.