%A OUYANG Yan-min, WANG Chang-feng, LIU Liu, YUAN Hong-min %T Study on the Risk Early Warning Interval Model Based on Improved Adaptive Optimal Partition Method——Large-scale Public Health Emergency %0 Journal Article %D 2022 %J Chinese Journal of Management Science %R 10.16381/j.cnki.issn1003-207x.2021.1327 %P 196-206 %V 30 %N 11 %U {http://www.zgglkx.com/CN/abstract/article_18216.shtml} %8 2022-11-20 %X Large-scale public health emergency, such as COVID-19, has severely endangered the safety of people around the world. Relevant emergency management system needs to be improved, among which risk early warning is the key issue. How to quantitatively estimate the risk early warning thresholds and partition risk early warning intervals is studied. In this paper, improved adaptive optimal partition model based on entropy method is introduced to fix this problem. It is found that the trend of public health emergency changes dynamically, while traditional partition model can not identify the changes of data characteristics in different epidemic periods. The model in this paper uses eight kinds of fitting functions to reduce the sum of squares of deviations when partitioning, so as to more accurately identify the development trends of epidemic in different periods. In addition, considering the diversity of epidemic indicators, five kinds of epidemic indicators are considered and the entropy method is used to assign the weight of each indicator, which can better avoid possible deviations caused by manual selection. In order to verify the effectiveness of the model, the data of daily COVID-19 cases worldwide from January 20, 2020 to March 31, 2020, are used and the simulation results of risk early warning intervals are given. It is found that our partition results obtained by improved adaptive optimal partition model are basically consistent with the WHO statements about the epidemic development stages. What’s more, comparing with traditional partition model, the results in this paper are more accurate and the risk early warning thresholds are more advanced. Therefore, the study has guiding value for enhancing the accuracy of risk early warning of public health emergencies, and provides a theoretical reference for the construction of emergency management system.