基于CEEMD - GRNN组合模型的HIV感染病例数预测

目的 构建基于百度指数的CEEMD - GRNN模型预测HIV感染病例数、为信息缺乏的HIV感染疫情预测提供可靠的方法，旨在为艾滋病流行趋势的传统预测方法提供有益补充。方法 第一，利用GRNN建立HIV感染病例数原始序列与百度指数的非线性关系；第二，先利用CEEMD提取HIV感染病例数的周期，再利用GRNN建立提取后序列与百度指数的非线性关系；第三，基于上述两种思想进一步建立组合预测模型，称为CEEMD - GRNN组合模型；最后，将CEEMD - GRNN组合模型应用于HIV感染病例数的预测。结果 模型拟合结果表明，最优单项模型的MAPE为10.17%，CEEMD - GRNN组合模型的MAPE为7.18%，组合模型的预测精度高于最优单项模型。结论 本文提出的CEEMD - GRNN组合模型预测精度优于最优单项模型，所提模型能够为信息不充足的非线性HIV感染病例数据提供稳定可靠的预测方法。

Prediction of HIV infection cases nased on CEEMD-GRNN model

Objective To construct a Complementary Ensemble Empirical Mode Decomposition-generalized regression neural network(CEEMD-GRNN) model based on the Baidu index to predict the number of HIV infection cases and provide a reliable method for HIV infection epidemic prediction without sufficient information, aiming to provide a useful supplement to the traditional prediction methods of HIV epidemic trends. Methods First, GRNN was used to establish the nonlinear relationship between the original sequence of HIV infection cases and Baidu index. Second, CEEMD was used to extract the period of the number of HIV infected cases, then GRNN was used to establish the nonlinear relationship between the extracted sequence and Baidu index. Thirdly, based on the above two ideas, the combined prediction model was further established, which was called the CEEMD-GRNN combined model. Finally, the CEEMD-GRNN combination model was applied to predict the number of HIV infection cases. Results The results of model fitting showed that the MAPE of the optimal single-item model was 10.17%, and that of the CEEMD-GRNN combination model was 7.18%. The prediction accuracy of the combined model was higher than that of the optimal single-item model. Conclusion The prediction accuracy of the CEEMD-GRNN combined model proposed in this paper is better than that of the optimal single model, and the proposed model can provide a stable and reliable prediction method for the nonlinear HIV infection case data with insufficient information.