基于LASSO回归模型与百度搜索数据构建的流感疫情预测系统

发布时间：2018-06-08 00:37

本文选题：Bagging + LASSO　；参考：《中国卫生统计》2017年02期

【摘要】：目的探讨基于LASSO回归模型与百度搜索数据构建流感疫情预测系统的可行性。方法采用Bagging方法和模型性能的多指标优化评估策略,对传统LASSO回归模型进行改进,构建性能提升的集成LASSO回归模型,将其应用于中国大陆地区季节性流感流行趋势的预测研究。结果与传统LASSO回归模型相比,本研究所构建的集成LASSO回归模型对2011年至2015年中国大陆地区流感流行趋势的预测偏差更小,说明集成LASSO回归模型的外部推断能力更强,适合于流感疫情的预测分析;本研究开发了开源的R软件程序包Sparse Learner,方便用户进行调用和进一步开发研究。结论 Bagging方法和多指标优化评估策略相结合所构建的集成LASSO回归模型,有效地增强了传统LASSO回归模型的性能。本研究所构建的预测模型可以应用于流感等传染病疫情的预测研究。
[Abstract]:Objective to explore the feasibility of constructing influenza epidemic forecasting system based on LASSO regression model and Baidu search data. Methods by using bagging method and multi-index optimization evaluation strategy of model performance, the traditional LASSO regression model was improved, and an integrated LASSO regression model with improved performance was constructed, which was applied to predict seasonal influenza epidemic trend in mainland China. Results compared with the traditional LASSO regression model, the integrated LASSO regression model constructed in this study has a lower prediction deviation to the influenza epidemic trend in mainland China from 2011 to 2015, which indicates that the integrated LASSO regression model has stronger external inference ability than the traditional LASSO regression model. It is suitable for the prediction and analysis of influenza epidemic. In this study, open source R software package Sparse Learneris developed, which is convenient for users to call and further research. Conclusion the integrated LASSO regression model based on bagging method and multi-index optimization evaluation strategy can effectively enhance the performance of the traditional LASSO regression model. The prediction model can be used to predict the epidemic situation of infectious diseases such as influenza.
【作者单位】：汕头大学医学院公共卫生与预防医学教研室;中山大学公共卫生学院医学统计与流行病学系;中山大学卫生信息中心、广东省卫生信息学重点实验室;
【分类号】：R511.7

【相似文献】