| 寻鲁宁,王 冲,沈成凤,等.1990—2019年中国结直肠癌发病趋势分析及预测模型比较[J].中国肿瘤,2023,32(4):279-286. |
| 1990—2019年中国结直肠癌发病趋势分析及预测模型比较 |
| Prediction of Colorectal Cancer Incidence in China with Three Different Models Based on Trends from 1990 to 2019 |
| 投稿时间：2022-12-13 |
| DOI：10.11735/j.issn.1004-0242.2023.04.A007 |
|
 |
| 中文关键词: 结直肠癌 趋势 ARIMA模型 GM(1，1)模型 XGBoost模型 预测 |
| 英文关键词:colorectal cancer trend ARIMA model GM(1, 1) model XGBoost model predict |
| 基金项目:天津市卫生高层次人才选拔培养项目（津人才［2018］19号）；天津市医学重点学科（专科）建设项目（TJYXZDXK-051A） |
|
| 摘要点击次数: 899 |
| 全文下载次数: 149 |
| 中文摘要: |
| 摘 要：［目的］ 分析中国结直肠癌发病趋势，建立并比较预测模型在结直肠癌标化发病率中的应用。［方法］ 收集全球健康数据交换（Global Health Data Exchange，GHDx）数据库中1990—2019年中国结直肠癌发病率数据，利用Joinpoint线性回归模型分析计算结直肠癌发病率的年度变化百分比（annual percentage change，APC）。运用R4.1.2软件分别构建ARIMA模型、灰色模型以及XGBoost模型，比较三个模型拟合预测效果，选取最优模型对2020—2025年中国结直肠癌标化发病率进行预测。［结果］ 1990—2019年中国结直肠癌发病率以6.19%的速度稳步上升（标化发病率APC=3.66%，P＜0.001），男性增速（发病率7.08%，标化发病率4.46%）高于女性（发病率4.93%，标化发病率2.50%）。构建的三个模型中，ARIMA(1，1，1)模型拟合预测效果优于GM(1，1)模型和XGBoost模型，相对误差（relative error，RE）在0.07%~2.26%之间，均方根误差（root mean square error，RMSE）、平均绝对误差（mean absolute error，MAE）和平均绝对百分比误差（mean absolute percentage error，MAPE）分别为0.286、0.231和1.134%。选取最优模型预测2020—2025年中国结直肠癌标化发病率，预测值依次为31.36/10万、32.05/10万、32.63/10万、33.14/10万、33.56/10万和33.93/10万。［结论］ 1990—2019年中国结直肠癌发病呈稳步上升趋势，ARIMA(1，1，1)模型在结直肠癌标化发病中具有良好的预测性能。 |
| 英文摘要: |
| Abstract: ［Purpose］ To predict the incidence of colorectal cancer in China with different forecasting models. ［Methods］ The incidence data of colorectal cancer in China from 1990 to 2019 were collected from global health data exchange(GHDx) database. Joinpoint linear regression model was used to calculate the annual percentage change(APC) for colorectal cancer. R4.1.2 software was used to establish ARIMA model， gray model and XGBoost model， respectively. By comparing the fitting and forecasting performance of three models， the optimal model was selected to predict age-standardized incidence rate (ASIR) of colorectal cancer in China from 2020 to 2025. ［Results］ From 1990 to 2019， the incidence of colorectal cancer in China increased steadily both in crude incidence rate and ASIR(APC=6.19% and 3.66%， respectively， P<0.001）， and the APCs of crude incidence rate and ASIR in male was higher than those in females（7.08% vs 4.93% and 4.46% vs 2.50%）. The ARIMA(1， 1， 1) model had better fitting and forecasting performance than GM(1， 1) model and XGBoost model. The relative error was between 0.07% and 2.26%. The root mean square error(RMSE)， mean absolute error(MAE) and mean absolute percentage error(MAPE) were 0.286， 0.231 and 1.134%， respectively. The predicted values of colorectal cancer ASIR from 2020 to 2025 were 31.36/105， 32.05/105， 32.63/105， 33.14/105， 33.56/105， and 33.93/105， respectively. ［Conclusion］ From 1990 to 2019， the incidence of colorectal cancer in China shows a steady upward trend， and the ARIMA(1， 1， 1) model has good performance in predicting the age-standardized incidence rate of colorectal cancer. |
|
在线阅读
查看全文 查看/发表评论 下载PDF阅读器 |