| 胡文斌,张 婷,秦 威.ARIMA乘积季节模型在恶性肿瘤月度发病率预测中的应用[J].中国肿瘤,2017,26(12):960-966. |
| ARIMA乘积季节模型在恶性肿瘤月度发病率预测中的应用 |
| Forecasting the Monthly Cancer Incidence Using the Autoregressive Integrated Moving Average Model |
| 投稿时间：2017-04-18 |
| DOI：10.11735/j.issn.1004-0242.2017.12.A007 |
|
 |
| 中文关键词: 季节性求和自回归移动平均模型 恶性肿瘤 X-12 时间序列 |
| 英文关键词:autoregressive integrated moving average model cancer X-12 time series |
| 基金项目:昆山市社会发展科技项目（KS1655） |
|
| 摘要点击次数: 1624 |
| 全文下载次数: 394 |
| 中文摘要: |
| 摘 要：［目的］ 基于昆山市2006~2015年月度恶性肿瘤发病率，采用X-12自回归移动平均混合模型（ARIMA）建立时间序列模型，并用模型预测未来年份恶性肿瘤发病率，以期为恶性肿瘤的防治工作提供指导。［方法］ 采用X-12-ARIMA季节调整乘积模型对昆山市2006年1月至2015年12月恶性肿瘤发病率进行趋势成分分解，并自动选择ARIMA季节调整乘积模型，以贝叶斯信息准则（BIC）值最小为最优模型选择标准；以绝对误差值、平均绝对百分比误差及决定系数（R2）来评价模型拟合精度。［结果］ X-12选择的最优乘积季节模型为ARIMA(0，1，1)×(0，1，1)12；预测方程为：(1-β)(1-β12)Xt =(1-0.84687β)(1-0.70249β12)εt。建模结果显示2006~2015年恶性肿瘤月度发病率不仅呈现季节波动，而且2007年1月至2018年12月发病率呈现持续上升的长期趋势。精度评价指标绝对误差值、平均绝对百分比误差及决定系数（R2）分别为2.13/10万、7.30%和0.697。［结论］ ARIMA(0，1，1)×(0，1，1)12能够应用于预测昆山市恶性肿瘤月度发病率，在过去10年恶性肿瘤发病率上升，并且未来年份继续上升的背景下，提示肿瘤相关疾病负担和未来健康保健资源将大幅增加。 |
| 英文摘要: |
| Abstract：［Purpose］ To forecast cancer incidence by time-series analysis with a Seasonal Autoregressive Integrated Moving Average Model(ARIMA). ［Methods］ Cancer incidence was obtained from Kunshan cancer registry from January 2005 to December 2016，and X-12-ARIMA was performed to automatic ARIMA model selection. Bayesian Information Criterion(BIC) was used to evaluate the statistics fitness. Absolute error，the mean absolute percentage error(MAPE) and determinant coefficients(R2) were used to examine the precision of model fitness. ［Results］ Based on the monthly cancer incidence between 2006 to 2015，the model ARIMA(0，1，1)×(0，1，1)12 had good fitness，and the forecasting equation was(1-β)(1-β12)Xt =(1-0.84687β)(1-0.70249β12)εt，the cancer incidence with moving average coefficient and seasonal moving average coefficient were statistically significant(P<0.01). The predicted monthly cancer incidence was consistent with the actual numbers between January to December in 2014 with a mean absolute error，MAPE and R2 of 2.13/105，7.30% and 0.697，respectively. ［Conclusion］ The ARIMA model(0，1，1)×(0，1，1)12 can be used to predict the monthly cancer incidence. The model shows that monthly cancer incidence has been increased for more than a decade and it will remain increasing in the near future in Kunshan city. |
|
在线阅读
查看全文 查看/发表评论 下载PDF阅读器 |