中国肾癌死亡趋势预测及其预测模型比较

目的  建立并比较两种预测模型在中国肾癌死亡趋势中的应用，选取最佳模型对2020—2025年中国肾癌死亡率进行预测。方法  收集全球健康数据交换(Global Health Data Exchange, GHDx)数据库1990—2019年中国人群全年龄组肾癌死亡率数据，使用R 4.0.2软件基于1990—2016年肾癌死亡率数据分别建立自回归移动平均模型(autoregressive integrated moving average model, ARIMA)和灰色模型(gray model, GM)(1, 1)，比较2017—2019年预测值与实际值以评价两种模型的拟合和预测性能，采用最佳模型预测2020—2025年中国肾癌死亡情况。结果  1990—2019年中国肾癌粗死亡率(crude mortality rate, CMR)呈上升趋势；在备选的ARIMA模型中，ARIMA(1, 2, 0)拟合效果最好，GM(1, 1)模型表达式为Y(t+1)=9.267 8e^{0.050 2(t)}-8.771 0；ARIMA(1, 2, 0)模型的平均绝对误差(mean absolute error, MAE)、均方根误差(root mean squared error, RMSE)和平均绝对百分比误差(mean absolute percent error, MAPE)在拟合部分和预测部分均低于GM(1, 1)模型；根据最佳模型预测结果，2025年中国肾癌死亡率相比于2019年将增加7.74%。结论  较于GM(1, 1)模型，ARIMA(1, 2, 0)模型对我国肾癌死亡率的拟合效果和预测性能更好。

Prediction of kidney cancer mortality trends and comparison of the two prediction models in China

  Objective  To build and compare two forecasting models for kidney cancer mortality in China, and choose the best model to predict the mortality rate of kidney cancer in China from 2020 to 2025.  Methods  We collected the mortality data of kidney cancer in China among all ages in the global health data exchange (GHDx) database from 1990 to 2019. R 4.0.2 software were used to establish autoregressive integrated moving average model (ARIMA) model and gray model (GM) (1, 1) model. Based on kidney cancer mortality data from 1990 to 2016, we compared the predicted and actual values from 2017 to 2019 to evaluate the fitting and forecasting performance of the two models. Finally, we selected the optimal model to predict kidney cancer mortality in China from 2020 to 2025.  Results  There was an increasing trend for the crude mortality rate (CMR) of kidney cancer in China from 1990 to 201. The ARIMA (1, 2, 0) model has the best fitting effect among those alternative ARIMA models. The model expression of GM (1, 1) was Y(t+1)=9.267 8e0.050 2(t)-8.771 0. The mean absolute error (MAE), root mean squared error (RMSE) and mean absolute percent error (MAPE) of ARIMA (1, 2, 0) were lower than GM (1, 1) on fitting and forecasting performance. According to the best model prediction results, the mortality of kidney cancer in China will increase by 7.74%in 2025 compared to 2019.  Conclusion  Compared with GM (1, 1) model, ARIMA (1, 2, 0) model had better fitting effect and forecasting performance for kidney cancer mortality in China.