key: cord-0975686-xhml87np
authors: Xiong, Li; Hu, Peiyang; Wang, Houcai
title: Establishment of epidemic early warning index system and optimization of infectious disease model: Analysis on monitoring data of public health emergencies
date: 2021-09-02
journal: Int J Disaster Risk Reduct
DOI: 10.1016/j.ijdrr.2021.102547
sha: 8dca4af8fd1a250c07d0d5decbceaf222f792baf
doc_id: 975686
cord_uid: xhml87np

The ability to mitigate the damages caused by emergencies is an important symbol of the modernization of an emergency capability. When responding to emergencies, government agencies and decision makers need more information sources to estimate the possible evolution of the disaster in a more efficient manner. In this paper, an optimization model for predicting the dynamic evolution of COVID-19 is presented by combining the propagation algorithm of system dynamics with the warning indicators. By adding new parameters and taking the country as the research object, the epidemic situation in countries such as China, Japan, Korea, the United States and the United Kingdom was simulated and predicted, the impact of prevention and control measures such as effective contact coefficient on the epidemic situation was analyzed, and the effective contact coefficient of the country was analyzed. The results revealed that China's Grade I Response to public health emergencies after the COVID-19 outbreak effectively prevented the persistent spread of new coronavirus, making the pandemic controlled in a relatively short period of time compared with other countries and regions. This study reaffirmed the importance of responding quickly to public health emergencies and formulating prevention and control policies to reduce population exposure and prevent the spread of the pandemic.

Public health emergency is an unexpected emergency event in the risk society, which causes great damage 26 to all aspects of social life. Public health emergencies are characterized by uncertainty, complexity, diffusion and 27 cross-domain [1] . Among them, the most typical is a major contagious disease event. Infectious diseases have 28 been a disease that severely threatens human life and health and seriously affects the development of social 29 economy [2-3]. According to a COVID-19 real time statistical data report by World Health Organization (WHO) 30

in August 1 st , 2021, revealed that the global cumulative numbers to 198.5 million cases and over 4.2 million 31 deaths cases since the start of the pandemic. The number of deaths from the COVID-19 pandemic [4] . According 32

In order to improve the SEIR model and make it more in line with the current status of COVID-19, we mine 23

the key information of the occurrence of public health emergencies, and establish a set of scientific, reasonable 24 and sensitive index system [9] . Predict the situation trend of infectious diseases based on huge amounts of 25 epidemic data and other data that have a serious impact on the pandemic. With the assistance of large data and 26 artificial intelligence, it is vital to quantify the influencing factors of the epidemic situation and to establish the 27 whole process evaluation index system on the basis of forming an effective epidemic data structure. It is crucial 28 to combine the theoretical prediction algorithm with the actual early warning index [10] . 29 According to the above research ideas, the rest of this paper is organized as follows: the second part describes 30 the existing relevant research, which is the cornerstone of the follow-up research. The third and fourth parts 31 respectively discussed Formulate Preventive Indicator System and Prediction Model based on SEIR separately. 32

The fifth part is a synthesis of the experiments carried out in the first two parts. In the end of the paper, it contains 33 pointed out that in addition to the common prevention and control measures, public awareness of the epidemic 23 also has a significant impact on the spread of COVID- 19. 24 At present, some countries have established a unified monitoring, warning and reporting system for public 25 health emergencies, like Electronic Surveillance System for Early Notification of Community-based Epidemics 26 (ESSENCE) in the US. However, the system did not respond in time in COVID-19, which may be due to the 27 single source of monitoring and early warning data indicators in the existing early warning system. The value of 28 a large number of spatio-temporal big data has not been mined and reflected, and the monitoring data still mainly 29 comes from medical and health institutions; The data information is one-sided and lacks other information of 30 great significance for early monitoring and early warning, such as specific symptoms, contact history, life history, 31 traffic history, etc. it is impossible to accurately carry out infectious disease early warning only by relying on the 32 data of clinical diagnosis results. 33 To sum up, a handful of studies have been the research basis for systematic research based on the data 34 compilation of the COVID-19 and related indicators. However, the research and practical application of COVID-35 19 are mostly based on static case data, and there is a lack of evaluation indicators to extract and summarize 36 various data. There are many factors influencing the outbreak of infectious diseases and they cover a wide range 37 of areas, the comprehensive support of multiple data is increasingly needed to improve the monitoring and early 38 warning capabilities. If we can set up a monitoring and early warning index system based on multivariate data 39 for infectious disease epidemic, and take multivariate data into account, we can further improve the accuracy and 40 sensitivity of early warning, so as to play a certain supporting role in the study of early warning decision-making 41 of new emergency think tanks, and strengthen and highlight their risk early warning capabilities. Therefore, it is 42 necessary to analyze and sort out the data of each link in detail to form an index system for epidemic assessment.

Jumping out of the background of COVID-19, many scholars have conducted machine learning-based 23 identification model analysis and research on various epidemics before. For example, in view of the lag of 24 traditional infectious disease surveillance, a hidden Markov model for epidemic identification and surveillance 25 is proposed, which combines large data from Internet search engine and WHO infectious disease surveillance 26 data. Using the time characteristics of the spread of infectious diseases in the same country or region, this method 27 can achieve a high accuracy and real-time identification and monitoring of infectious disease outbreaks in a single 28 country or region. Compared with traditional methods, the lag is much reduced [26] . Combining basic data with 29 other factors, Pan et al. [27] proposed to use Long Short-Term Memory (LSTM) loop neural network to predict 30 the incidence of infectious diseases, in order to effectively improve the accuracy of the identification model. 31

Indeed, many researchers are starting to carry out data-based epidemic prediction studies using various 32 algorithms to supplement the deficiencies of existing prediction systems. In these studies, large data, such as 33

Internet search queries, are being used abroad to monitor the occurrence of infectious diseases. Internet search 34 data can be collected and processed at a near real-time speed. Sherry et al. [28] have found that searching for data 35 over the Internet can create infectious disease identification models faster than traditional monitoring systems. In 36 addition, some scholars such as Huang et al. [29] have attempted to use the generalized additive model (GAM) 37 to identify and predict hand, foot and mouth disease, which includes the best results obtained by searching and 38 querying data to obtain new tools for identifying and monitoring large data, which have the advantage of easy 39 access to the incidence of infectious diseases and can identify infectious disease trends before official 40 organizations. 41 the relationship between social factors such as personnel mobility, population density and specific prevention and 1 control measures. Different parameters have a great impact on the prediction results, so it is necessary to 2 accurately introduce relevant parameters and complete the prediction. This model has strong pertinence, but its 3 generalization ability needs to be further improved. The dynamic prediction research based on machine learning 4 does not need to consider too many influencing factors, but its results are directly affected by the data correlation. 5

For short-term prediction, small errors can be obtained by using relevant machine learning methods, but with the 6 passage of time, the errors will gradually accumulate. Without other intervention measures, the prediction may 7 deviate from the actual results seriously, and the sustainability is not strong [30-32]. The next part in this paper 8 will dig several kinds of data to understand the evolution status after the outbreak, and gain insight into the risk 9 level of the current epidemic gap. On this basis, determine the influencing factors of the epidemic situation and 10 quantify them, establish a scientific and reasonable index system, establish a corresponding system dynamic 11 model, and predict the model based on the index system. To analyze the impact of national emergency 12 management policies on the trend of COVID- 19. 13 

Feature engineering is a term used in the field of machine learning [33]. In the era of big data, data has 16 become so complex that it needs to accurately capture its most essential characteristics. In other words, the feature 17 is to extract useful information from the dataset to predict the results. It is also the main basis for predicting 18 samples of machine learning algorithm models and an important factor to determine the upper performance limit 19 of algorithm models. In datasets with different structures, there are different forms of features [34] . In a tabular 20 (structured) dataset consisting of variables or attributes (columns) and instances (rows), variables or attributes 21 can be considered characteristics. In a nutshell, features are the information that is processed from the original 22 data to describe the individual samples. Before features are actually used for prediction, there is no guarantee that 23 the original attributes or extracted features will have a positive effect on the result prediction [35] . To sum up, in 24 order to get a practical and effective prediction model, this paper first needs to make an index system to ensure 25 that the extracted data features are effective, which is called feature engineering. The flow diagram is shown in 26 generating some information describing the characteristics of the sample, replacing the original data as input to 1 the model, and making the features play a better role in the algorithm [36] . In the field of computer science, there 2 is the saying that "feature engineering determines the upper limit of generalization ability". Thus, feature 3 engineering is important throughout the model process. Owing to a difficult and time-consuming process, feature 4 engineering requires a variety of possibilities tests. In the actual data mining work, it takes a lot of time to 5 complete feature extraction and organization, and then only a simple model is needed to obtain good prediction 6 results [37] . 7

For early warning, a set of index system must be established to analyze and evaluate the situation of an 9 infectious disease, judge the probability and severity of the crisis, and decide whether to send a crisis alarm. Early (2) Early-symptom-period indicators. These indicators refer to that when the alarm source indicators change 7 abnormally, there are always some omens, mainly manifested in the increase in the sales of all kinds of over-the-8 counter drugs that may be related to infectious diseases in the hospital and the increase in patients with similar 9 clinical symptoms with infectious diseases, which can be regarded as the warning indicators of atypical symptoms. 10 strong outbreak, etc. It also interacts with the public events themselves and constantly affects the security risks 20 of public events. Therefore, public opinion indicators also play a supplementary role. 21

Based on the basic structure of infectious disease epidemic warning indicators system, this paper combines 22 the process of the occurrence and development of new coronary pneumonia with the basic principles of early 23 warning theory, collects and sorts out various possible data collections, and proposes a three-level basic 24 framework of COVID-19 warning indicator system (Table 1) . 25 The construction of this indicator system covers the whole process of the occurrence and development of 2 infectious diseases, and comprehensively considers the multifaceted data such as pathogen surveillance data, 3 natural outbreak data, hospital reporting data, social and network media, which can fully mine the value of big 4 data and improve the timeliness and accuracy of infectious disease information warning. Meanwhile, based on 5 the classification of early warning indicators, this index system is divided into three levels to form a whole, which 6 has a certain theoretical basis. Among the foregoing warning indicators, some can be obtained by publishing data 7 sources; some can be obtained unconventionally, but can also be obtained through some efforts in the current 8 situation; and a few others are difficult to obtain under current conditions, but remain in the indicator system due 9

to the importance of indicators for early warning work. 10

Because geographical variations, economic differences, and other factors, different regions, different 11

infectious diseases have their own epidemic characteristics, so for all regions, a uniform system of early warning 12 indicators should be established for all diseases, and it is inappropriate and impossible to set a fixed cordon. The 13

indicator system in this paper is built for infectious diseases, especially COVID- 19. There is an overall and 14 individual relationship when the specific application, and the early warning indicators should be rounded and 15 revised according to the characteristics of specific diseases and specific regions. 16 In the process of epidemic development, a large number of data produced by the epidemic show the 17 characteristics of multi-source, massive, rapid growth and constant updating. Therefore, the construction of early 18

warning index system by integrating multi-party monitoring data is undoubtedly a further promotion and 19 supplement to the existing early warning system and existing research. Its analysis process and results also 20 provide a certain theoretical basis for the next part of the research on epidemic prediction using SEIR model. 21

The prediction of infectious diseases can early detect the trend of disease development, lay the foundation 23 for the early warning of disease, and provide a theoretical basis for formulating prevention strategies and 24 measures. When responding to the COVID-19 pandemic, it is of great significance to establish an appropriate 25 prediction model and improve the accuracy of prediction. 26

System dynamics, also known as industrial dynamics, mainly refers to the use of system simulation to 28 analyze and solve management problems [52]. The study of system dynamics on system problems is based on 29 the close dependence of the internal behavior mode and system structure of the system. By establishing a 30 mathematical model, the causal relationship in changing forms can be gradually discovered [53-54]. As the study 31 of System Dynamics have been applied to the study of various fields. By combining various research methods, 32 more accurate mathematical simulation results can be obtained, which provides a basis for decision-making. At 33 present, system dynamics is increasingly being used in socioeconomic research to explore economic and social 34 operating patterns through system design and simulation, thus providing managers with more reliable 35 recommendations [55]. System dynamics methods have also been applied to the study of major contagious 36 diseases. through system construction and simulation, the relationship between indicators in the system can be parameters, and functional relationships, which can be used to analyze the relationship between factors and 1 dynamic change trends. According to the traditional SEIR model, the population within the epidemic scope of infectious diseases is 8 divided into four categories: , namely susceptible population, refers to the population who has not yet acquired 9 the disease, but lacks immunity and is susceptible to infection after contacting the susceptible population; , 10 namely exposed population, refers to the susceptible population who initially experiences the incubation period 11 and will have symptoms after a period of time; refers to the population who is infected with infectious diseases, relatively relaxed conditions mainly through differential variance. However, it may not be directly used in the 1 prediction of COVID- 19. In the first step of modifying the model, we first need to analyze the characteristics of 2 COVID- 19 . Unlike some viruses that caused infection in the past, such as SARS, which show infectivity only 3 during the outbreak period, the COVID-19 is infectious in the latent period. Based on these characteristics, we The above flowchart shows the evolution of COVID-19 without human intervention. Accordingly,, it is 9 necessary to determine the infection probability of the 2 model, the propagation probability of the model is 10 modified, so that the susceptible individuals is transformed into the latent, and the number of healthy susceptible 11 contacted by the latent every day is 2 ( ). 12 Therefore, on the basis of formula (1), it is necessary to add latent E besides ( ) and ( ) to S infection 13 probability 2 and contact number 2 ( ) of susceptible healthy people, which will make more normal people In addition to the characteristics of COVID-19, the intervention of human environment has a significant 19

impact on the trend of the epidemic. In the third part of the study, we have mentioned the epidemic situation 20 identification indicators, which is the main object of study in this part. In addition, we have summarized indicators of potential risk factors. The data represented in this section is likely to be an influential factor in the SEIR model, 1

interfering with the direction of prediction. The main transmission routes of SARS-CoV-2 are respiratory foam 2 transmission and close contact transmission, with contacts and infected persons becoming the main source of 3

Therefore, based on the SEIR model flow chart, combined with the potential risk factor indicators in the 5 previous part, this paper obtains three new factors: the number of isolation susceptible subjects, the number of 6 suspected cases and the number of confirmed cases, so as to represent the change of epidemic situation after 7 intervention by human factors. The flow chart of the optimized model is shown in Fig.7 According to Fig.7 , the differential equation of the SEIR model is also modified. Assuming that the total 12 number of persons managing a region is constant, ( ) = ( is a constant). Let the isolation proportion be , 13 the contact rate be , and the effective contact coefficient be (1 is taken in the natural state), then be the 14 effective contact rate. The conversion rate from susceptible to isolated susceptible was (1 − ) , from 15 susceptible to isolated latent was , and from susceptible to latent was (1 − ). Considering the impact 16 of infected persons without isolation I and latent persons E on susceptible population S, and that SQ will become 17 susceptible again after de-isolation, we choose to indicate the rate of de-isolation, = 1/day of isolation . 18 Record that the rate of latent transformation to the infected person is , = 1/latent days, the isolation rate of 19 the infected person is , the mortality rate is 1, and the cure rate is 2. The optimized formula (3) In the above-mentioned model, part of the index data comes from the epidemic surveillance data certified 23 by China National Health and Construction Commission and the official notification, which ensures the reliability of the data on the one hand, and the authenticity of the model on the other hand. Secondly, the weight of the new 1 impact parameters is affected by the actual emergency management mechanism of each country. The policy 2 constant is determined by the expert scoring method, and the index weight is calculated by the analytic hierarchy 3 process to ensure the accuracy of the prediction results. 4

Based on the above COVID-19 epidemic early warning system, the basic data of SEIR model was obtained Table  18 2. 19 in other countries, indicating that epidemic prevention and control measures in the remaining few countries have 28 played a role. In this study, it may not be meaningful to directly predict epidemic inflection points in these 29

countries, but these true data can still be used to verify the feasibility and validity of the models in this study. 

The development of the epidemic is closely related to the selection of values. In the SEIR model, β represents 20 the conversion rate from exposure to disease, which is usually determined by the nature of the disease itself. disease is around 7 days, so β=1/7. In addition, other parameters, such as probability of infection σ, isolation ratio 23 q, contact rate c, recovery rate γ, are closely related to the specific social environment and prevention and control 24 policies of each country. In view of this, we use the data set of the first ten days of each country for parameter 25 fitting and error testing, and generates the best historical simulation state under a series of parameter sequences 26 through continuous trial and error, so as to alleviate the problem of inaccurate prediction model caused by 27 parameters. Among them, we set at the beginning that if there is no protection policy, the contact rate in the area 28 is 1, and the isolation ratio does not exist. In addition, we adopted the view of other experts that "the infection 29 probability β of susceptible persons after contacting patients is about 0.19 to 0.24, which may be slightly different 30 From the lockdown of Wuhan on January 23, 2020, to the semi-closed state of major epidemic prevention cities, 3 people across the country stayed at home and rested in line with national policies. Enterprises across the country 4 were suspended, schools were suspended, and the entire society is in a state of shutdown, and the contact rate 5 will decrease accordingly. Finally, the values of q and γ are related to the current medical conditions of various 6 countries. Based on the indicators sorted out in the third part, we set different values under different actual 7 conditions in various countries, and through continuous trial and error, we adjusted the parameters in order to 8 make more realistic, reasonable and standardized forecasts. 9 When the number of new diagnoses decreases, the inflection point is reached. When the number of new 10 cases increases to zero, the cumulative number of infected cases reaches the maximum, which is the peak of the 11 cumulative number of infected persons. In this paper, the optimized SEIR model equation is converted into a 12 program, and according to the model, real data from different countries are introduced to test the prediction results, 13

which are integrated as shown in Table 4 . 14 

According to Table 4 , the pandemic in China, Korea, and Japan can be effectively controlled before April 17 30, 2020, and new cases will gradually show a decreasing trend. Among them, China's inflection point has the 18 earliest appearance date, with a turn after a large outbreak of about one and a half months. Additional projections 19

show that the United States is the country with the largest number of patients in the world, and even in general 20 model projections, the highest peak in the number of patients in the United States is not seen and can only be 21 further speculated on the basis of projections toward. Apart from that, the situation in the UK with respect to the 22 pandemic situation trend is also somewhat less optimistic, with the number of patients predicted to reach its peak 23 in Mid-June, 2020, approximately 333 000 cases being diagnosed cumulatively, at a time when nearly four 24 months have passed since the pandemic just started. Based on this, United Kingdom and United States still need 25 severe prevention and control, and even more proper emergency management measures are needed to prevent 26 further epidemic spread and warn citizens of every effort to avoid travel to countries with similar conditions such 27 as United Kingdom and the United States. Citizens in these areas should also avoid out aggregation and take self-28 protection measures, paying attention to wearing masks. 29

As of the end of August 2021, the global impact of COVID-19 has not yet ended, and even due to the sudden 31 mutation of the virus, SARS-CoV-2, it has become more difficult to cope with the epidemic. Based on the 32 previous analysis, we believe that the conclusions of this study are important in order to bring warnings to the 33 policies of various countries. 34

The main content of this paper includes the following two aspects: Firstly, based on the level of early 35 warning indicators, five analysis dimensions of the indicators of the nature of infectious diseases, indicators of has been established. Second, it is based on the SEIR model in system dynamics, adjusting parameters according 2 to the policies of various countries, fitting research on the dynamic progress, and simulating and predicting the 3 future evolution trend of COVID-19 epidemic. In the experiment, this paper focuses on sorting out the value 4 logic of the impact factors based on the indicator system, and through continuous trial and error, to obtain a 5 predictive model with a higher degree of fit, and analyzes COVID-19 under the premise of different emergency 6 measures in various countries. The parameters of the prediction model are set in combination with various 7 indicators. The prediction trend can be adjusted through a variety of parameters to obtain results with higher 8 credibility than traditional models. Therefore, when choosing an infectious disease prediction model, we should 9 consider the characteristics and transmission routes of the epidemic disease in many ways, as well as the 10 intervention measures of the prevention and control department. 11 Therefore, the early warning index system and optimization of infectious disease model proposed in this 12 paper is an effective method, which can predict the peak and inflection point of this round of epidemic situation 13 and the corresponding total number of infections many days in advance. By adjusting the parameters, it can also 14 judge the approximate date of the infection peak. This will help the government to better plan their emergency 15 management means and prevention and control measures. 16

Based on the research in this paper, we believe that the forecast and early warning of major infectious 18 diseases represented by COVID-19 need to include two aspects of research simultaneously. First of all, it is 19 necessary for us to find sensitive and effective early warning indicators as well as establish a complete early 20

warning monitoring system. On the one hand, an indicator system can be used to collect and filter useful and 21 relevant data from the existing massive and responsible information to ensure that the data is timely, complete 22 and accurate. On the other hand, we also hope that through the guidance of an indicator system, we will have the 23 opportunity to discover some new valuable data, which is valuable for early detection of infectious disease 24 outbreaks or epidemics, or for more accurate prediction and warning. In addition, at least at this stage, there is no 25 mathematical model that can be directly applied to the prediction and early warning of various infectious diseases. 26

Although many models have a certain degree of generalization. However, in practical applications, different 27 infectious viruses have their own characteristics, and the prediction model must also be modified accordingly to 28 enhance the ability of prediction and early warning . 29 The paper strives to provide early warning of emergencies scientifically and effectively through the 30 combination of these two technologies, and put forward feasible references for the implementation of various 31 countermeasures. Judging from the conclusion, the paper has made certain contributions to the prediction and 32 early warning research of infectious disease epidemics on both theoretical and practical levels. 33 Firstly, in terms of theory, the paper provides new research perspectives and methods for epidemic forecasting 34 and early warning research. Monitoring and early warning of infectious diseases based on multi-source data 35

analysis is an inevitable trend in the development of the information age, and most of the existing studies in this 36 field have relatively single data sources. Since the epidemic of infectious diseases is characterized by population, 37 time and spatial distribution, it is a concentrated expression of the interaction of natural geographical 38 environmental factors, social and economic development conditions, and the flow of people in a specific time 39 and space. During the development of the epidemic, its occurrence A large amount of data presents the 40 characteristics of multi-source, massive, rapid growth, and constantly updated. Therefore, the paper constructs 41 an early warning indicator system by integrating multi-party monitoring data and applies it to the construction of 42 predictive models. Through the establishment of more accurate models, the overall utilization of epidemic data Secondly, in terms of reality, the paper puts forward a new idea based on big data thinking to construct the 2 forecast and early warning of infectious disease epidemics, which can be effectively applied to actual prevention 3 and control. The early warning indicators in this paper are based on the analysis of comprehensive big data, which 4 can be regarded as the data structure of sudden public health incidents. Through the collection and analysis of 5 data on meteorological factors, population flow trends, and the incidence of various syndromes, etc., Study the 6 external factors and individual internal factors that affect the occurrence and development of emergency 7 management events. Although the SEIR model selected in this paper can better capture the historical changes of 8 the epidemic in actual forecasting, and can reveal the autocorrelation of the development of the epidemic, it also 9 shows that the model is affected by various factors such as the incubation period of the disease, prevention and 10 control policies and measures. The impact of is relatively large, so if you use this model alone to make predictions, 11 you need to carefully study the relevant factors, and the setting of parameters will affect the accuracy of the 12 prediction. The combination of indicators and forecasts has also improved the actual operability of the two aspects 13 to a certain extent. 14 This paper combines the forecasting ideas and comprehensively analyzes the data of COVID-19, and obtains 15 more comprehensive results. However, there is still room for further improvement in research. First of all, the 16 index system proposed in this paper is based on the various data and information that has attracted attention after 17 the occurrence of COVID-19, and has been extracted using feature engineering methods, but there is still a lack 18 of actual verification. The development and improvement of the index system must be explored and verified in 19 practice, so as to further remove the indicators with low early warning value or difficult to obtain in practice, 20 classify and simplify the indicators with high relevance, improve the operability of the index system as much as 21 possible and promote the application of the index system. Therefore, it is necessary to carry out pilot studies to 22 gradually explore the application of early warning indicators. The evaluation and verification of indicators 23 requires the accumulation of a certain amount of historical data. Due to the time limit of writing this paper, there 24

is not enough data to verify and evaluate the indicators. Second, early warning and forecasting are based on 25 historical data and actual materials to predict the future, and provide suggestions for the management department 26 to grasp the status quo and the future. Prediction is only an analysis of the future development trend of infectious 27 diseases, but early warning is different from prediction. It needs to alert early abnormal situations and initiate 28 emergency response. This thesis only stays in the forecasting stage. How to combine forecasting and early 29 warning in depth is the direction and focus of further development. Finally, the occurrence and development of 30 infectious diseases are affected by many factors. If we can observe the dynamic process of a group of (multi-31 dimensional) related early warning indicators at the same time, and study it as a whole, and use appropriate 32 models to predict, try to integrate multiple indicators and multiple methods for analysis, it is bound to be able to 33

More systematically and comprehensively reflect the internal regularity and future trends of dynamic phenomena. 34 This will also be our future research thinking . 35 In conclusion, the combination of index system and prediction method for epidemic surveillance is an 36 alternative research direction and path, which provides a feasible way to study the spread of infectious diseases. 

Emotional "inflection point" in public health emergencies with 6 the 2019 new coronavirus pneumonia (ncp) in china

Prevalence of comorbidities in individuals diagnosed and undiagnosed with alzheimer's disease in león

spain and a proposal for contingency procedures to follow in the case of emergencies involving people with alzheimer's disease. 10

How scientific research reacts to international 12 public health emergencies: a global analysis of response patterns

Sars-cov-2, sars-cov-1 and mers-cov viral load 14 dynamics, duration of viral shedding and infectiousness -a living systematic review and meta-analysis. The Lancet Microbe

Public health emergencies and human rights: problematic jurisprudence arising from the 16 COVID-19 pandemic

Innovation in the time of covid-19

Using health information exchange to 19 improve public health

The covid-19 infodemic. The Lancet Infectious Diseases

A spatiotemporal analysis of urban resilience to the covid-19 pandemic in the yangtze river delta

Comparing the accuracy of several network-24 based covid-19 prediction algorithms -sciencedirect

Perceived impact of the Covid-19 crisis on SMEs in different industry sectors

COVID-19: a promising cure for the global panic

The bright and dark sides of social media usage 30 during the covid-19 pandemic: Survey evidence from Japan

People-centered early warning systems in china: a bibliometric analysis 32 of policy documents

A big data-driven dynamic estimation model of relief 34 supplies demand in urban flood disaster

Clinical features of patients infected with 2019 novel 36 coronavirus in wuhan, china. The Lancet

Population flow drives spatio-temporal distribution 38 of COVID-19 in china

Twitter for disaster relief through sentiment 40 analysis for covid-19 and natural hazard crises

Predicting COVID-19 using hybrid ai model. reliability of online information on covid-19

Mathematical modelling of sir for COVID-19 forecasting

Prediction modelling of covid-19 outbreak in indonesia using a logistic regression model

Management strategies in a seir model of covid-19 community spread

Developing a COVID-19 mortality risk prediction 8 model when individual-level data are not available

Artificial intelligence (ai) 10 and big data for coronavirus (COVID-19) pandemic: a survey on the state-of-the-arts

Workpiece Quality Prediction Research Based on Multi-source Heterogeneous Industrial Big Data

Prediction of alzheimer's disease based on bidirectional lstm

Mass media and the 16 contagion of fear: the case of ebola in america

Monitoring hand, foot and mouth disease by combining search engine query 18 data and meteorological factors

Identifying the predictors of covid-19 infection outcomes and development of prediction 20 models

Real-time tracking and prediction of covid-19 infection using digital 22 proxies of population mobility and mixing

Prediction-based analysis on power consumption gap 24 under long-term emergency: a case in china under covid-19

Feb-stacking and feb-dnn models for stock trend prediction: a performance analysis 26 for pre and post covid-19 periods

Eleven routine clinical features predict covid-19 28 severity uncovered by machine learning of longitudinal measurements

A systematic review of automated feature engineering solutions in machine 30 learning problems. SBSI'20: XVI Brazilian Symposium on Information Systems

A conceptual basis for feature engineering

Feature selection: a data 34 perspective

Diagnostic serial interval as a novel indicator for contact tracing 36 effectiveness exemplified with the sars-cov-2/covid-19 outbreak in south korea

Framework for evaluating public health surveillance systems for 39 early detection of outbreaks

Effect of non-pharmaceutical interventions to 41 contain covid-19 in china

Disease outbreak detection 3 system using syndromic data in the greater washington dc area 1

Clinical features of patients infected with 2019 novel coronavirus 5 in wuhan, china. The Lancet

Smell and taste dysfunction in 7 patients with covid-19: a systematic review and meta-analysis

Strategic spatiotemporal vaccine distribution increases the survival rate in an 9 infectious disease like covid-19

Proposal of the autopsy guideline for infectious diseases: preparation for 11 the post-covid-19 era

A population-based cohort study of 13 socio-demographic risk factors for covid-19 deaths in sweden

Risk factors for mortality 15 of coronavirus disease-2019 (covid-19) patients in two centers of hubei province, china: a retrospective analysis

The pca-seq method applied to analyze of the 18 dynamics of covid-19 epidemic indicators

Positive externalities of an epidemic: the case of the corona virus 20 (covid-19) in china

Covid-19 effect on consumer payment systems: establishing cause & effect 23 relationship

Teaching strategic thinking using System Dynamics: lessons from a strategic development course

System dynamics modelling of urbanization under energy 27 constraints in china

Management strategies in a seir-type model of covid 19 community 29 spread

Nowcasting and forecasting the potential domestic and international spread of 31 the 2019-nCoV outbreak originating in Wuhan, China: a modelling study

The effectiveness of quarantine of wuhan city against the 33 corona virus disease 2019 (covid-19): a well-mixed seir model analysis

Accounting for Symptomatic and 35

Asymptomatic in a SEIR-type model of COVID-19

A new seaird pandemic prediction 37 model with clinical and epidemiological data analysis on covid-19 outbreak isolated people in incubation

Forecast and analysis of covid-19 epidemic based on improved seir model

Mathematical Modeling for Prediction Dynamics of the Coronavirus impact digital contact tracing effectiveness for covid-19

Forecasting the Worldwide Spread of COVID-19 based on 3 Logistic Model and SEIR Model

Incubation period of covid-19: a meta-analysis

Incubation Period of COVID-19 From 11545 Patients in 13 Observation Study

Clinical characteristics of coronavirus disease 2019 15 in china

We acknowledge the supports from the National Social Science Foundation of China (No. 21ZDA105). 2 3 4