CN101894309A - Epidemic situation predicting and early warning method of infectious diseases - Google Patents

Epidemic situation predicting and early warning method of infectious diseases Download PDF

Info

Publication number
CN101894309A
CN101894309A CN2009101853439A CN200910185343A CN101894309A CN 101894309 A CN101894309 A CN 101894309A CN 2009101853439 A CN2009101853439 A CN 2009101853439A CN 200910185343 A CN200910185343 A CN 200910185343A CN 101894309 A CN101894309 A CN 101894309A
Authority
CN
China
Prior art keywords
early warning
data
information
symptom
sequence
Prior art date
Application number
CN2009101853439A
Other languages
Chinese (zh)
Inventor
彭志行
Original Assignee
南京医科大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 南京医科大学 filed Critical 南京医科大学
Priority to CN2009101853439A priority Critical patent/CN101894309A/en
Publication of CN101894309A publication Critical patent/CN101894309A/en

Links

Abstract

The invention discloses an epidemic situation predicting and early warning method of infectious diseases, particularly a network early warning system of relevant public health events of infectious diseases in countries, provinces, cities (regions) and districts (counties). The method comprises the steps of integrating adverse reaction information data which are collected by the server side, calculating and determining the early warning threshold of public health events according to the quantities of the symptoms of adverse reactions on the basis of the operation of a time sequence analysis statistic model and automatically carrying out early warning in real time at any time. The method comprises the following specific steps of: (1) integrating information from different channels; (2) executing a path I: analyzing tasks; and (3) executing a path II: predicting and early warning. Once the quantity of monitoring data meets the traditional basic requirements of a tome sequence analysis method for considering the periodicity, the seasonal nature, long term tend and the like of diseases, the conclusion having relative-high value can be obtained to play important roles of early finding epidemic situations, taking measures as soon as possible, preventing the spread of the epidemic situations and preliminarily establishing an epidemic situation monitoring and early warning system and a working mechanism of the province.

Description

Epidemic situation predicting and early warning method of infectious diseases

Technical field

The present invention is the network early warning system of a kind of country, province, city (area), the relevant public health event of district's (county) aspect infectious disease, belongs to the technical field of epidemic prediction early warning.

Background technology

For can help government comprehensively, science, timely to effective early warning of carrying out of public health emergency, native system is integrated the information from different channels, and the relevant public health event of infectious disease is predicted that the theoretical research result of early warning technology and method is applied in the public health monitoring.

Each medical service of unit accurately uploads information datas such as unit basic condition information, employee's basic condition, report of infectious disease case in real time to data center, data center is the processing enter of provincial health supervision data sink bulk analysis, by form and icon form, synthesis the whole province health supervision operational data.

For the monitoring of unit epidemic situation in various degree research is arranged, comprise the sales volume of medical service of unit (school) non-prescribed medicine is carried out statistical study, the information of classification of drug is monitored, judge and the situation of breaking out of the employee of the unit of prediction (school student) epidemic with this; Carry out prediction and the analysis that employee's infectious disease is broken out epidemic situation according to unit (school) in the Monitoring Data of worker (student) aspect due to illness absent from duty.

By analyzing as can be known, the sales volume of medical service of unit non-prescribed medicine and the information of classification of drug are difficult to comprehensively, and covering employee's (school student) epidemic breaks out situation, worker (student) Monitoring Data due to illness absent from duty is also only accomplished detailed classification, could reflect the campus epidemic more accurately, because at present the time series of Monitoring Data often has interruption (between winter and summer vacation is arranged), periodicity for the infectious disease of data reflections, characteristics such as seasonality also can't be carried out very comprehensive convictive analysis and prediction, so alternative analytical approach is less.And, can't go can not find epidemic situation from bigger scope by grading forewarning system by analytical approachs such as time series analyses now to uploading to data of database analysis.

Summary of the invention

Technical matters: thus for can to the data that obtained better analyze and early warning at any time from wider discovery epidemic situation, the invention provides a kind of epidemic situation predicting and early warning method of infectious diseases, in case the amount of Monitoring Data reaches the basic demand of traditional consideration disease cycle, seasonality, secular trend equal time sequence analysis method, just can draw more valuable relatively conclusion, for we find epidemic situation early, take measures as early as possible, prevent spreading of epidemic situation, and tentatively set up our province epidemic monitoring early warning system and working mechanism plays a significant role.

Technical scheme: epidemic situation predicting and early warning method of infectious diseases of the present invention is made integration to the bad reaction information data that server end gathers, computing based on the time series analysis statistical model, symptom quantity according to bad reaction, the threshold value of warning of calculating and definite Public Health Emergencies, at any time a situation arises carries out real-time early warning to infectious disease automatically, and concrete grammar is as follows:

1). integrate information, be specially from different channels:

Automatically the database before refreshing before each execution early warning task in time adds to the information before the pre-warning time in the database; After obtaining data, will carry out by two paths: the path 1.: analysis; The path is 2.: with the time series analysis model serves as that the basis statistics is drawn coordinate diagram, and then carries out prediction and early warning task;

2). execution route 1.: analysis task:

Automatically generate not bad reaction statistical form and the statistical graph of commensurate, area, sex, time period in the database after the information via statistical study;

The information that utilization counts is tested to the model that following a period of time epidemic situation situation is predicted to definite being used in " prediction " task,

Because needed jiggly time series is converted into stationary sequence before definite time series models, this process is called the pre-service of data; And the stationarity of data sequence is meant that the mean value of this sequence and variance are always constant;

Set up corresponding coordinate diagram during deal with data, data are showed on coordinate diagram, link up with curve then, thereby the equation of being set up is called match with the consistent process that reaches the accurate purpose of prediction of information that chart is reflected as best one can;

Concrete grammar is:

I. historical epidemic situation generation data in software picked at random a period of time are that object is set up corresponding time series models with a part wherein; With an other part is that standard comes the model prediction effect is estimated,

The data that obtain among the II I are through pre-service, make it to help model is discerned and decide rank, and then definite required model;

The model of determining among the III.II carries out match to data sequence pretreated among the II and makes prediction to following a period of time incidence, check the degree of agreement of predicted value and actual value by on same coordinate diagram, making incidence figure and morbidity prognostic chart, and then this model accuracy is estimated;

3). execution route 2.: the prediction early warning

The time series analysis model is to be used for to object to be processed, and just the model added up of data based on the time series analysis model, carries out analytic statistics and is depicted as coordinate diagram the current data through after integrating of early warning of will predicting;

31). enter prediction module

A. judge according to the data plot and the autocorrelation function of the data sequence of integrating out whether the time series that is obtained is stationary sequence, if then need not handle, if not then need make it be processed into stationary sequence; Concrete grammar is:

If a1. sequences y is linear trend, the average instability is then utilized first order difference;

A2. present secondary trend as infructescence, average is not a constant, then utilizes second order difference;

A3. present in time rising or decline deviation as infructescence, variance is not a constant, then can utilize natural logarithm with its tranquilization usually,

B. the identification of model and decide rank

The identification of model: the product seaconal model is combining of seaconal model and RIMA model at random, pure ARIMA on the statistics (q) the model note is done for p, d:

Φ ( B ) ▿ d X t = Θ ( B ) ϵ t

Wherein, t represents the time, and Xt represents response sequence, and B is a backward shift operator, ▿ = 1 - B , P, d, q represent autoregression exponent number, difference order and moving average exponent number respectively; Φ (B) expression autoregression operator; Θ (B) expression running mean operator,

Exponent number be (p, d, q) * (P, D, Q) SThe product seaconal model can show be:

Φ ( B ) U ( B S ) ▿ d ▿ S D X t = V ( B S ) Θ ( B ) ϵ t

ε tRepresent and independently scratch or stochastic error, the value of s is the number of observation in the circulation in season, Expression is with the correlationship of different cycles point in one-period, Then described the correlationship on the corresponding time point in the different cycles, the two combines the effect of just having portrayed two factors simultaneously;

Adopt exponent number identification and the parameter estimation of Box2Jenkins method to the product seaconal model, just be based on investigating sample auto-correlation, the partial correlation function of data, season, length S was obtained by the analysis of actual application background; If sample auto-correlation, partial correlation function neither truncation do not trail yet, and not to be linear attenuation trend, on the contrary, on integral multiple point corresponding to cycle S, auto-correlation or partial correlation function the sizable peak value of absolute value occurs and present change in oscillation, just can the judgment data sequence be suitable for the product seaconal model;

Model decide rank: adopt the way of souning out choose difference and season difference order d, D; D, D should get lower-order, gets 1,2 or 3 usually; If for a certain group of d, auto-correlation that D obtains or partial correlation function present truncation preferably or hangover characteristic, then think corresponding d, and D suits, at this moment if increase d, and D, then corresponding auto-correlation or partial correlation function can present discrete the increase and non-steady state; As exponent number d, after D determines, model parameter p, q, P, the estimation of Q is generally with maximal possibility estimation and no constraint least square;

C sets up corresponding model in view of the above through discerning and deciding after the rank,

D. the effect assessment of model prediction,

Whether the goodness of fit by testing model decides model suitable

White noise refers to power spectrum density equally distributed noise in whole frequency domain, and the power of white noise signal on each frequency range is the same, and white noise is more convenient on mathematics manipulation, is the strong instrument of systematic analysis therefore;

If with a model fitting data, then the difference of Model Calculation value and data measured value is a residual error;

Computing machine is analyzed the residual error of observed reading and model fitting value, if residual sequence is not a white noise sequence, illustrate that then also having information to be included in the relevant residual sequence is not extracted, other parameters of model can not be represented the statistical property of modeling object fully, be that institute's established model is not a final mask, this moment can be to residual error match complicated model more, with the information of abundant refinement data, thereby obtain more suitably model

If residual sequence is not a white noise sequence, then need rebulid model, repeat above-mentioned steps, till residual sequence is white noise sequence;

32). software enters the early warning module

32a. calculate the mean value and the standard deviation of historical data automatically according to chart, be designated as μ and σ respectively; And calculate the value of μ+2 σ, μ+3 σ, μ+6 σ automatically,

32b. determine grade scale

If μ+2 σ are yellow early warning standard

μ+3 σ are orange early warning standard

μ+6 σ are the red early warning standard

The corresponding number of certain early warning symptom is designated as symptom quantity in the early warning unit,

This method is carried out grading forewarning system with threshold value as the corresponding actual symptoms quantity of standard,

Symptom quantity<μ+2 ρ then refuse early warning

μ+2 ρ<symptom quantity<μ+3 ρ then are yellow early warning

μ+3 ρ<symptom quantity<μ+6 ρ then are orange early warning

Symptom quantity>μ+6 ρ then are red early warning,

32c. unit, symptom, quantity information according to statistics calculate early warning information,

The epidemic information of μ that utilization is calculated and σ and the actual symptoms that counts and corresponding symptom quantity carries out grading forewarning system according to the grade scale calculating early warning information of front,

32d. the automatic adjustment of threshold value:

Bad reaction information before each early warning in the automatic refresh can farthest conform to data with actual feelings, and the calculated threshold of drawing thus, and threshold value is changed with varying environment; In case of emergency, also can be by data base administrator's mandatory modification threshold value,

32e. show early warning information through output on computers after the aforementioned calculation processing, comprise Alert Level, early warning symptom and quantity, Alert Level branch wherein: yellow early warning, orange early warning and red early warning, the early warning symptom is personnel's bad reaction symptom that server obtained, and quantity has the number of this symptom for this unit.Information can show automatically or be clicked by any unit in the monitoring range and obtain,

32f. the early warning on wider: calculate according to time series analysis early warning information to this unit early warning after, look concrete condition on this basis the epidemic situation early warning made in wider area.

Beneficial effect: methods such as operate time sequential analysis are reasonably analyzed the data that database obtains, and obtain in real time, result accurately, with regional characteristics and season dynamically adjusting threshold value of warning, guarantee the science of early warning.And, in case certain unit (school) has found the trend that has epidemic situation to break out, city/district/county the Disease Control and Prevention Center at this unit (school) place and the staff at provincial control center can receive the warning message that system sends automatically behind login system, and view the concrete personnel component of due to illness scarce duty and the concrete symptom of bad reaction, thereby take measures rapidly.

And, each unit (school) does not all reach the early warning number that due to illness lacks duty (class) in certain district/county, but during due to illness the lacking duty (class) number and reached the early warning number in district/at county level of the whole district/county, system also can remind the city/district/county Disease Control and Prevention Center that lands and the staff at provincial control center automatically, and sends alerting signal.Native system has stronger statistical study and predictive ability, no matter be that due to illness the unit (school) of basic unit lacks duty (class) and monitor the responsible official, or the staff of province/city/district/county Disease Control and Prevention Center, login system enters after the link of " office worker (student) statistics ", and due to illness employee's (campus student) that all can see arbitrary period under the aspect separately lacks the chart-information of duty (class) statistical graph, statistical form, sequence chart and other hommizations of enriching and to the judgement of future trend.And, along with system is more and more longer in the time of use in future, after historical data constantly increases in the system, use the effect of these statistical methods more can be protected, constantly adjusting the thresholding that draws through computer program in the future can more and more realistic requirements of one's work.

Description of drawings

Fig. 1 is a schematic flow sheet of the present invention.

Embodiment

1 server obtains information

Information spinner will comprise

1, student's number that misses classes on account of illness every day: as long as miss classes on account of illness every day promptly counted miss classes on account of illness in 1 class hour, count 1 people.

2, student's rate that misses classes on account of illness every day: the number that the misses classes on account of illness/number of should turning out for work.

3, student's fate that misses classes on account of illness: the person of missing classes on account of illness counts one day more than 4 hours, not enough meter half a day.

4, monitoring of diseases syndrome: heating, cough, headache, pharyngalgia, stomachache, diarrhoea, vomiting, blood-shot eye illness, fash, injured, other.

5, monitoring of diseases kind: flu, tracheitis, pneumonia, varicella, rubella, measles, mumps, enterogastric diseases, heart disease, illness in eye, odontopathy, otolaryngologic disease, urologic disease, neurasthenia, unexpected injury, tuberculosis, hepatitis, other infectious diseases, agnogenio disease.

6, miss classes on account of illness cardinal symptom or disease reason people number of times: a student surpasses 1 day because of same disease is absent from school, and symptom and disease are all counted 1 person-time.

2 softwares are integrated information, mainly are to refresh original database automatically, and the school's bad reaction information before the pre-warning time is in time added in the database.Owing to epidemic situation a situation arises can be along with different seasons, different places present different characteristics, so must with up-to-date, can represent at that time that the epidemic situation generation data of actual conditions are object, so just can make correct prediction, correct early warning.

Information after 3 integration can be used for analyzing.

Automatically generate different schools, class, sex, bad reaction statistical form and the statistical graph of time period after the bad reaction information via statistical study of school;

These chart-information data of can be used as are used for definite being used in " prediction " task tested to the model that following a period of time school's epidemic situation situation is predicted.

Concrete grammar is:

I. the historical epidemic situation generation of the school in software picked at random a period of time data are that object is set up corresponding time series models with a part wherein; With an other part is that standard comes the model prediction effect is estimated,

The data that obtain among the II I are through pre-service, make it to help model is discerned and decide rank, and then definite required model;

The model of determining among the III.II carries out match to data sequence pretreated among the II and makes prediction to following a period of time school's incidence, check the degree of agreement of predicted value and actual value by on same coordinate diagram, making incidence figure and morbidity prognostic chart, and then this model accuracy is estimated

4 based on the time series analysis model, and the current epidemic situation generation data through after integrating of this school of early warning of will predicting are carried out analytic statistics and are depicted as coordinate diagram

(1) carries out the prediction task: promptly utilize available data to pass through the processing of this software to following this school's epidemic situation prediction that a situation arises of a period of time

A school bad reaction information is showing on coordinate diagram after the software processes, judge according to data sequence data plot and autocorrelation function on this figure whether the time series that is obtained is stationary sequence, if then need not handle, if not then need make it be processed into stationary sequence; Concrete grammar is:

If a1. sequences y is linear trend, the average instability is then utilized first order difference;

A2. present secondary trend as infructescence, average is not a constant, then utilizes second order difference;

A3. present in time rising or decline deviation as infructescence, variance is not a constant, then can utilize natural logarithm with its tranquilization usually,

The identification of b model and decide rank

The identification of model: the product seaconal model is combining of seaconal model and RIMA model at random, pure ARIMA on the statistics (q) the model note is done for p, d:

Φ ( B ) ▿ d X t = Θ ( B ) ϵ t

Wherein, t represents the time, and Xt represents response sequence, and B is a backward shift operator, ▿ = 1 - B , P, d, q represent autoregression exponent number, difference order and moving average exponent number respectively; Φ (B) expression autoregression operator; Θ (B) expression running mean operator,

Exponent number be (p, d, q) * (P, D, Q) SThe product seaconal model can show be:

Φ ( B ) U ( B S ) ▿ d ▿ S D X t = V ( B S ) Θ ( B ) ϵ t

ε tRepresent and independently scratch or stochastic error, the value of s is the number of observation in the circulation in season, Expression is with the correlationship of different cycles point in one-period, Then described the correlationship on the corresponding time point in the different cycles, the two combines the effect of just having portrayed two factors simultaneously;

Adopt exponent number identification and the parameter estimation of Box2Jenkins method to the product seaconal model, just be based on investigating sample auto-correlation, the partial correlation function of data, season, length S was obtained by the analysis of actual application background;

If sample auto-correlation, partial correlation function neither truncation do not trail yet, and not to be linear attenuation trend, on the contrary, on integral multiple point corresponding to cycle S, auto-correlation or partial correlation function the sizable peak value of absolute value occurs and present change in oscillation, just can the judgment data sequence be suitable for the product seaconal model;

Model decide rank: adopt the way of souning out choose difference and season difference order d, D; D, D should get lower-order, gets 1,2 or 3 usually; If for a certain group of d, auto-correlation that D obtains or partial correlation function present truncation preferably or hangover characteristic, then think corresponding d, and D suits, at this moment if increase d, and D, then corresponding auto-correlation or partial correlation function can present discrete the increase and non-steady state; As exponent number d, after D determines, model parameter p, q, P, the estimation of Q is generally with maximal possibility estimation and no constraint least square;

C sets up corresponding model in view of the above through discerning and deciding after the rank,

D. the effect assessment of model prediction,

Whether the goodness of fit by testing model decides model suitable, that is to say with utilize that this model draws a situation arises predicts the outcome and existing data are expressed on same coordinate diagram by the processing of this software to the following a period of time epidemic situation of school, come whereby both residual errors are analyzed.

Computing machine is analyzed the residual error of observed reading and model fitting value, if residual sequence is not a white noise sequence, illustrate that then also having information to be included in the relevant residual sequence is not extracted, other parameters of model can not be represented the statistical property of modeling object fully, be that institute's established model is not a final mask, this moment can be to residual error match complicated model more, with the information of abundant refinement data, thereby obtain more suitably model

If residual sequence is not a white noise sequence, then need rebulid model, repeat above-mentioned steps, till residual sequence is white noise sequence;

(2) carry out the early warning task

The number that campus student misses classes on account of illness all meets or approximately meets a time series at the incidence of each different time sections.In case the amount of Monitoring Data reaches the basic demand of traditional consideration disease cycle, seasonality, secular trend equal time sequence analysis method, just can draw more valuable relatively conclusion, for we find epidemic situation early, take measures as early as possible, prevent spreading of epidemic situation

A. calculate the mean value and the standard deviation of school's epidemic situation data according to chart automatically, be designated as μ and σ respectively; And calculate the value of μ+2 σ, μ+3 σ, μ+6 σ automatically,

B. determine grade scale

If μ+2 σ are yellow early warning standard

μ+3 σ are orange early warning standard

μ+6 σ are the red early warning standard

There is the number of student of this kind early warning symptom to be designated as symptom quantity in the school,

This method is carried out grading forewarning system with threshold value as the corresponding actual symptoms quantity of standard,

Symptom quantity<μ+2 ρ then refuse early warning

μ+2 ρ<symptom quantity<μ+3 ρ then are yellow early warning

μ+3 ρ<symptom quantity<μ+6 ρ then are orange early warning

Symptom quantity>μ+6 ρ then are red early warning,

C. according to school, symptom, ill number of student information calculations early warning information,

The epidemic information of μ that utilization is calculated and σ and the actual symptoms that counts and corresponding symptom quantity carries out grading forewarning system according to the grade scale calculating early warning information of front to school,

D. the automatic adjustment of threshold value:

School's bad reaction information before each early warning in the automatic refresh can farthest conform to data with actual feelings, and the calculated threshold of drawing thus, and threshold value is changed with varying environment; In case of emergency, also can be by data base administrator's mandatory modification threshold value.The software control threshold value the school of different seasons, different scales, class dynamically the oneself adjust, must reach accurate early warning and provide safeguard for better

E. output on computers shows early warning information after handling through aforementioned calculation, comprise Alert Level, early warning symptom and quantity, Alert Level branch wherein: yellow early warning, orange early warning and red early warning, the early warning symptom is student's bad reaction symptom that server obtained, and quantity is the number of student that this symptom is arranged in the school.Information can show automatically or click acquisition by the personnel of medical service of school in the monitoring range,

F. the early warning on wider: each school does not all reach the early warning number that misses classes on account of illness in certain district/county, but when the number that misses classes on account of illness in the whole district/county has reached the early warning number in district/at county level, city/district/county the Disease Control and Prevention Center that the automatic prompting of system's meeting is landed and the staff at provincial control center, and send alerting signal

G Disease Control and Prevention Center proposes handling suggestion according to early warning situation and the fortuitous event that system monitoring arrives, and the monitoring school that will dispatch officers in case of necessity carries out student's morbidity survey, prevention and the control disease of taking measures diffusion and development.

Province Disease Control and Prevention Center exercises supervision to the various places disposition according to system information and instructs.

Claims (1)

1. epidemic situation predicting and early warning method of infectious diseases, it is characterized in that this method makes integration to the bad reaction information data that server end gathers, computing based on the time series analysis statistical model, symptom quantity according to bad reaction, the threshold value of warning of calculating and definite Public Health Emergencies, at any time a situation arises carries out real-time early warning to infectious disease automatically, and concrete grammar is as follows:
1). integrate information, be specially from different channels:
Automatically the database before refreshing before each execution early warning task in time adds to the information before the pre-warning time in the database; After obtaining data, will carry out by two paths: the path 1.: analysis; The path is 2.: with the time series analysis model serves as that the basis statistics is drawn coordinate diagram, and then carries out prediction and early warning task;
2). execution route 1.: analysis task:
Automatically generate not bad reaction statistical form and the statistical graph of commensurate, area, sex, time period in the database after the information via statistical study;
The information that utilization counts is tested to the model that following a period of time epidemic situation situation is predicted to definite being used in " prediction " task,
Because needed jiggly time series is converted into stationary sequence before definite time series models, this process is called the pre-service of data; And the stationarity of data sequence is meant that the mean value of this sequence and variance are always constant;
Set up corresponding coordinate diagram during deal with data, data are showed on coordinate diagram, link up with curve then, thereby the equation of being set up is called match with the consistent process that reaches the accurate purpose of prediction of information that chart is reflected as best one can;
Concrete grammar is:
I. historical epidemic situation generation data in software picked at random a period of time are that object is set up corresponding time series models with a part wherein; With an other part is that standard comes the model prediction effect is estimated,
The data that obtain among the III are through pre-service, make it to help model is discerned and decide rank, and then definite required model;
The model of determining among the III.II carries out match to data sequence pretreated among the II and makes prediction to following a period of time incidence, check the degree of agreement of predicted value and actual value by on same coordinate diagram, making incidence figure and morbidity prognostic chart, and then this model accuracy is estimated;
3). execution route 2.: the prediction early warning
The time series analysis model is to be used for to object to be processed, and just the model added up of data based on the time series analysis model, carries out analytic statistics and is depicted as coordinate diagram the current data through after integrating of early warning of will predicting;
31). enter prediction module
A. judge according to the data plot and the autocorrelation function of the data sequence of integrating out whether the time series that is obtained is stationary sequence, if then need not handle, if not then need make it be processed into stationary sequence; Concrete grammar is:
If a1. sequences y is linear trend, the average instability is then utilized first order difference;
A2. present secondary trend as infructescence, average is not a constant, then utilizes second order difference;
A3. present in time rising or decline deviation as infructescence, variance is not a constant, then can utilize natural logarithm with its tranquilization usually,
B. the identification of model and decide rank
The identification of model: the product seaconal model is combining of seaconal model and RIMA model at random, pure ARIMA on the statistics (q) the model note is done for p, d:
Φ ( B ) = ▿ d X t = Θ ( B ) ϵ t
Wherein, t represents the time, and Xt represents response sequence, and B is a backward shift operator, ▿ = 1 - B , P, d, q represent autoregression exponent number, difference order and moving average exponent number respectively; Φ (B) expression autoregression operator; Θ (B) expression running mean operator,
Exponent number be (p, d, q) * (P, D, Q) the product seaconal model of s can show be:
Φ ( B ) U ( B S ) ▿ d ▿ S D X t = V ( B S ) Θ ( B ) ϵ t
ε tRepresent and independently scratch or stochastic error, the value of s is the number of observation in the circulation in season, Expression is with the correlationship of different cycles point in one-period, Then described the correlationship on the corresponding time point in the different cycles, the two combines the effect of just having portrayed two factors simultaneously;
Adopt exponent number identification and the parameter estimation of Box2Jenkins method to the product seaconal model, just be based on investigating sample auto-correlation, the partial correlation function of data, season, length S was obtained by the analysis of actual application background; If sample auto-correlation, partial correlation function neither truncation do not trail yet, and not to be linear attenuation trend, on the contrary, on integral multiple point corresponding to cycle S, auto-correlation or partial correlation function the sizable peak value of absolute value occurs and present change in oscillation, just can the judgment data sequence be suitable for the product seaconal model;
Model decide rank: adopt the way of souning out choose difference and season difference order d, D; D, D should get lower-order, gets 1,2 or 3 usually; If for a certain group of d, auto-correlation that D obtains or partial correlation function present truncation preferably or hangover characteristic, then think corresponding d, and D suits, at this moment if increase d, and D, then corresponding auto-correlation or partial correlation function can present discrete the increase and non-steady state; As exponent number d, after D determines, model parameter p, q, P, the estimation of Q is generally with maximal possibility estimation and no constraint least square;
C sets up corresponding model in view of the above through discerning and deciding after the rank,
D. the effect assessment of model prediction,
Whether the goodness of fit by testing model decides model suitable
White noise refers to power spectrum density equally distributed noise in whole frequency domain, and the power of white noise signal on each frequency range is the same, and white noise is more convenient on mathematics manipulation, is the strong instrument of systematic analysis therefore;
If with a model fitting data, then the difference of Model Calculation value and data measured value is a residual error;
Computing machine is analyzed the residual error of observed reading and model fitting value, if residual sequence is not a white noise sequence, illustrate that then also having information to be included in the relevant residual sequence is not extracted, other parameters of model can not be represented the statistical property of modeling object fully, be that institute's established model is not a final mask, this moment can be to residual error match complicated model more, with the information of abundant refinement data, thereby obtain more suitably model
If residual sequence is not a white noise sequence, then need rebulid model, repeat above-mentioned steps, till residual sequence is white noise sequence;
32). software enters the early warning module
32a. calculate the mean value and the standard deviation of historical data automatically according to chart, be designated as μ and σ respectively; And calculate the value of μ+2 σ, μ+3 σ, μ+6 σ automatically,
32b. determine grade scale
If μ+2 σ are yellow early warning standard
μ+3 σ are orange early warning standard
μ+6 σ are the red early warning standard
The corresponding number of certain early warning symptom is designated as symptom quantity in the early warning unit,
This method is carried out grading forewarning system with threshold value as the corresponding actual symptoms quantity of standard,
Symptom quantity<μ+2 ρ then refuse early warning
μ+2 ρ<symptom quantity<μ+3 ρ then are yellow early warning
μ+3 ρ<symptom quantity<μ+6 ρ then are orange early warning
Symptom quantity>μ+6 ρ then are red early warning,
32c. unit, symptom, quantity information according to statistics calculate early warning information,
The epidemic information of μ that utilization is calculated and σ and the actual symptoms that counts and corresponding symptom quantity carries out grading forewarning system according to the grade scale calculating early warning information of front,
32d. the automatic adjustment of threshold value:
Bad reaction information before each early warning in the automatic refresh can farthest conform to data with actual feelings, and the calculated threshold of drawing thus, and threshold value is changed with varying environment; In case of emergency, also can be by data base administrator's mandatory modification threshold value,
32e. show early warning information through output on computers after the aforementioned calculation processing, comprise Alert Level, early warning symptom and quantity, Alert Level branch wherein: yellow early warning, orange early warning and red early warning, the early warning symptom is personnel's bad reaction symptom that server obtained, and quantity has the number of this symptom for this unit.Information can show automatically or be clicked by any unit in the monitoring range and obtain,
32f. the early warning on wider: calculate according to time series analysis early warning information to this unit early warning after, look concrete condition on this basis the epidemic situation early warning made in wider area.
CN2009101853439A 2009-11-05 2009-11-05 Epidemic situation predicting and early warning method of infectious diseases CN101894309A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2009101853439A CN101894309A (en) 2009-11-05 2009-11-05 Epidemic situation predicting and early warning method of infectious diseases

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2009101853439A CN101894309A (en) 2009-11-05 2009-11-05 Epidemic situation predicting and early warning method of infectious diseases

Publications (1)

Publication Number Publication Date
CN101894309A true CN101894309A (en) 2010-11-24

Family

ID=43103496

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2009101853439A CN101894309A (en) 2009-11-05 2009-11-05 Epidemic situation predicting and early warning method of infectious diseases

Country Status (1)

Country Link
CN (1) CN101894309A (en)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102222163A (en) * 2011-05-25 2011-10-19 中国人民解放军防化指挥工程学院 Epidemic situation prediction method for infectious disease having immune period and influenced by seasons
CN104504465A (en) * 2014-12-16 2015-04-08 国电南京自动化股份有限公司 Power generation fuel supply prediction method
CN104615884A (en) * 2015-02-04 2015-05-13 中国科学院地理科学与资源研究所 Severe infectious disease and mortality risk early warning system and method based on virus detection rate
CN106339569A (en) * 2015-07-09 2017-01-18 深圳迈瑞生物医疗电子股份有限公司 Method and a device for determining abnormality of sample test result
CN106960124A (en) * 2017-03-17 2017-07-18 北京农信互联科技有限公司 Livestock epidemic situation alarming method and device
CN107229828A (en) * 2017-05-24 2017-10-03 广州金域医学检验中心有限公司 The data early warning method and device tested and analyzed for endocrine
CN107239654A (en) * 2017-05-24 2017-10-10 广州金域医学检验中心有限公司 Data early warning method and device for prenatal and postnatal care examination
CN107330283A (en) * 2017-07-06 2017-11-07 江苏省疾病预防控制中心 A kind of method for early warning and device
CN107480452A (en) * 2017-08-17 2017-12-15 深圳先进技术研究院 Multi-user's mood monitoring method, device, equipment and storage medium
CN107945879A (en) * 2017-12-11 2018-04-20 创业软件股份有限公司 A kind of medical data method for digging for preventiveing treatment of disease
CN110111903A (en) * 2019-04-28 2019-08-09 中国疾病预防控制中心 A kind of method and apparatus of public health emergency prediction
WO2019153596A1 (en) * 2018-02-07 2019-08-15 平安科技(深圳)有限公司 Chicken pox incidence warning method, server, and computer readable storage medium
WO2019196281A1 (en) * 2018-04-11 2019-10-17 平安科技(深圳)有限公司 Epidemic grading and prediction method and apparatus, computer apparatus, and readable storage medium
CN111681774A (en) * 2020-08-11 2020-09-18 南京云联数科科技有限公司 Methods, computing devices, and media for epidemic prediction

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102222163A (en) * 2011-05-25 2011-10-19 中国人民解放军防化指挥工程学院 Epidemic situation prediction method for infectious disease having immune period and influenced by seasons
CN102222163B (en) * 2011-05-25 2016-11-09 中国人民解放军防化指挥工程学院 A kind of have duration of immunity and the epidemic Forecasting Methodology by seasonal effect
CN104504465A (en) * 2014-12-16 2015-04-08 国电南京自动化股份有限公司 Power generation fuel supply prediction method
CN104615884A (en) * 2015-02-04 2015-05-13 中国科学院地理科学与资源研究所 Severe infectious disease and mortality risk early warning system and method based on virus detection rate
CN104615884B (en) * 2015-02-04 2017-07-28 中国科学院地理科学与资源研究所 A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method
CN106339569A (en) * 2015-07-09 2017-01-18 深圳迈瑞生物医疗电子股份有限公司 Method and a device for determining abnormality of sample test result
CN106960124A (en) * 2017-03-17 2017-07-18 北京农信互联科技有限公司 Livestock epidemic situation alarming method and device
CN107229828A (en) * 2017-05-24 2017-10-03 广州金域医学检验中心有限公司 The data early warning method and device tested and analyzed for endocrine
CN107239654A (en) * 2017-05-24 2017-10-10 广州金域医学检验中心有限公司 Data early warning method and device for prenatal and postnatal care examination
CN107330283A (en) * 2017-07-06 2017-11-07 江苏省疾病预防控制中心 A kind of method for early warning and device
CN107480452A (en) * 2017-08-17 2017-12-15 深圳先进技术研究院 Multi-user's mood monitoring method, device, equipment and storage medium
CN107945879A (en) * 2017-12-11 2018-04-20 创业软件股份有限公司 A kind of medical data method for digging for preventiveing treatment of disease
WO2019153596A1 (en) * 2018-02-07 2019-08-15 平安科技(深圳)有限公司 Chicken pox incidence warning method, server, and computer readable storage medium
WO2019196281A1 (en) * 2018-04-11 2019-10-17 平安科技(深圳)有限公司 Epidemic grading and prediction method and apparatus, computer apparatus, and readable storage medium
CN110111903A (en) * 2019-04-28 2019-08-09 中国疾病预防控制中心 A kind of method and apparatus of public health emergency prediction
CN111681774A (en) * 2020-08-11 2020-09-18 南京云联数科科技有限公司 Methods, computing devices, and media for epidemic prediction
CN111681774B (en) * 2020-08-11 2020-11-27 南京云联数科科技有限公司 Methods, computing devices, and media for epidemic prediction

Similar Documents

Publication Publication Date Title
Zhang et al. Regional frequency analysis of droughts in China: a multivariate perspective
Grace et al. Fast processing of diel oxygen curves: Estimating stream metabolism with BASE (BA yesian S ingle‐station E stimation)
Kalra et al. Using large‐scale climatic patterns for improving long lead time streamflow forecasts for Gunnison and San Juan River Basins
Hernández et al. A study of the relationship between weather variables and electric power demand inside a smart grid/smart world framework
Holtgrieve et al. Simultaneous quantification of aquatic ecosystem metabolism and reaeration using a Bayesian statistical model of oxygen dynamics
Arnold et al. SWAT: Model use, calibration, and validation
Dise et al. Predicting dissolved inorganic nitrogen leaching in European forests using two independent databases
Nusser et al. The National Resources Inventory: a long-term multi-resource monitoring programme
Barth From child abuse to permanency planning: Child welfare services pathways and placements
Sivakumar Dominant processes concept, model simplification and classification framework in catchment hydrology
Wu et al. Artificial neural networks for forecasting watershed runoff and stream flows
Harmel et al. Estimating storm discharge and water quality data uncertainty: a software tool for monitoring and modeling applications
Jones et al. Surrogate Measures for Providing High Frequency Estimates of Total Suspended Solids and Total Phosphorus Concentrations 1
Perlin et al. Distribution of industrial air emissions by income and race in the United States: an approach using the toxic release inventory
Strobl et al. Network design for water quality monitoring of surface freshwaters: A review
Hu Spatial analysis of MODIS aerosol optical depth, PM 2.5, and chronic coronary heart disease
Wood et al. Prospects for advancing drought understanding, monitoring, and prediction
Briggs The role of GIS: coping with space (and time) in air pollution exposure assessment
Gillis et al. Monitoring Canada's forests: the national forest inventory
Watson et al. Currents connecting communities: nearshore community similarity and ocean circulation
Böse et al. CISN ShakeAlert: An earthquake early warning demonstration system for California
CN102282552B (en) Based Intelligent Control based on pattern, monitoring and the system of automatization, method and computer program
Borrego et al. Procedures for estimation of modelling uncertainty in air quality assessment
Hu et al. Particulate air pollution and chronic ischemic heart disease in the eastern United States: a county level ecological study using satellite aerosol data
D JOHNSON A brief history of the analysis of crime concentration

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20101124

C12 Rejection of a patent application after its publication