CN111524611A

CN111524611A - Method, device and equipment for constructing infectious disease trend prediction model

Info

Publication number: CN111524611A
Application number: CN202010334815.9A
Authority: CN
Inventors: 孙继超; 张子恒; 陈曦; 刘华罗; 郑冶枫
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-04-24
Filing date: 2020-04-24
Publication date: 2020-08-11
Anticipated expiration: 2040-04-24
Also published as: CN111524611B

Abstract

The invention provides a method for constructing an infectious disease trend prediction model, an epidemic situation trend prediction method, a device, electronic equipment and a computer readable storage medium; the method comprises the following steps: attenuating the basic infection number of the infectious diseases according to the transmission time to obtain effective infection numbers of a plurality of dates in the transmission period; determining fitting state data of a plurality of dates in one-to-one correspondence with effective infection numbers of the dates in a state conversion relation included in the infectious disease trend prediction model; extracting fitting case data of a plurality of dates from the fitting state data of the plurality of dates; and updating parameters of the infectious disease trend prediction model according to the difference between the real case data of the plurality of dates and the fitting case data of the plurality of dates, and taking the updated parameters as parameters used for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model. According to the epidemic situation trend prediction method and device, the infectious disease data can be combined to accurately model so as to support epidemic situation trend prediction based on the infectious disease trend prediction model.

Description

Method, device and equipment for constructing infectious disease trend prediction model

Technical Field

The invention relates to the field of intelligent medical treatment based on artificial intelligence technology, in particular to a method for constructing an infectious disease trend prediction model, an epidemic situation trend prediction method, an epidemic situation trend prediction device, electronic equipment and a computer readable storage medium.

Background

Artificial Intelligence (AI) is a comprehensive technique in computer science, and by studying the design principles and implementation methods of various intelligent machines, the machines have the functions of perception, reasoning and decision making. The artificial intelligence technology is a comprehensive subject and relates to a wide range of fields, for example, natural language processing technology and machine learning/deep learning, etc., and along with the development of the technology, the artificial intelligence technology can be applied in more fields and can play more and more important values.

The artificial intelligence technology is applied to infectious disease prediction, fitting case data can be obtained through the reasoning and decision functions of a machine, so that the epidemic situation trend of the infectious disease is predicted, and subsequent prevention and control measures are carried out according to the epidemic situation trend of the infectious disease. However, there is a lack in the related art of an effective scheme for infectious disease modeling and prediction based on artificial intelligence.

Disclosure of Invention

The embodiment of the invention provides a method for constructing an infectious disease trend prediction model, an infectious disease epidemic situation trend prediction method based on the infectious disease trend prediction model, a device, electronic equipment and a computer readable storage medium, which can be combined with infectious disease data for accurate modeling to support epidemic situation trend prediction based on the infectious disease trend prediction model.

The technical scheme of the embodiment of the invention is realized as follows:

the embodiment of the invention provides a method for constructing an infectious disease trend prediction model, wherein the infectious disease trend prediction model is used for predicting epidemic situation trend of infectious diseases; the method comprises the following steps:

attenuating the basic infection number of the infectious disease according to the transmission time to obtain effective infection numbers of a plurality of dates in the transmission period;

determining fitted state data of a plurality of dates in one-to-one correspondence with effective infection numbers of the plurality of dates in the state conversion relation of the infectious diseases included in the infectious disease trend prediction model;

extracting fitting case data of a plurality of dates from the fitting state data of the plurality of dates;

updating parameters of the infectious disease trend prediction model according to differences between real case data of a plurality of dates and fitted case data of the plurality of dates, and

and taking the updated parameters as parameters used for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model.

The embodiment of the invention provides a device for constructing an infectious disease trend prediction model, wherein the infectious disease trend prediction model is used for predicting epidemic situation trend of infectious diseases; the device comprises:

the attenuation module is used for attenuating the basic infection number of the infectious diseases according to the transmission time to obtain the effective infection numbers of a plurality of dates in the transmission period;

the determination module is used for determining fitting state data of a plurality of dates in one-to-one correspondence with the effective infection numbers of the dates in the state conversion relation of the infectious diseases included in the infectious disease trend prediction model;

the extraction module is used for extracting fitting case data of the dates from the fitting state data of the dates;

and the updating module is used for updating the parameters of the infectious disease trend prediction model according to the difference between the real case data of the plurality of dates and the fitting case data of the plurality of dates, and taking the updated parameters as the parameters used for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model.

In the above technical solution, the attenuation module is further configured to add the expected effective infection counts at the end of the infectious disease and the attenuation results corresponding to the basic infection counts at the plurality of dates to obtain the effective infection counts at the plurality of dates, respectively;

wherein the attenuation result is obtained by attenuating the basic infection number based on the transmission time of the infectious disease.

In the above technical solution, the attenuation module is further configured to determine an effective infection count of any one of the dates t according to the following formula:

R_t＝R_∞+R₀*e^-θ*twherein R is_tRepresents the effective infection number, R, of said date t₀*e^-θ*tRepresenting the result of the decay of said primary infection count at said date t, R₀Representing said base infection number, theta representing the rate of decline of said effective infection number, R_∞Indicating the expected effective number of infections at the end of the infection.

In the above technical solution, the determining module is further configured to determine, as fitting status data of the plurality of dates, status data that corresponds to the plurality of dates one by one and satisfies the constraint condition, with the state transition relationship of the infectious disease as a constraint condition and the effective infectious numbers of the plurality of dates as known quantities in the constraint condition;

wherein, the state transition relation of the infectious disease comprises parameters of the infectious disease trend prediction model.

In the above technical solution, the states of the infectious diseases include a susceptible state, an exposed state, an infected state and an isolated state;

the infectious disease trend prediction model comprises the state transition relationships including:

a conversion relationship between the susceptible state of date t, the infectious state of date t, an infection rate of the infectious disease, and a rate of decline of the susceptible state of date t,

a conversion relationship between the susceptible state of the date t, the infectious state of the date t, the exposed state of the date t, the infection rate of the infectious disease, the incidence rate of the infectious disease, and the rising rate of the exposed state of the date t,

a conversion relationship between the exposure state of the date t, the infection state of the date t, the incidence rate of the infectious disease, the quarantine rate of the infectious disease, and the rising rate of the infection state of the date t,

a conversion relationship between the infection state of the date t, an isolation speed of the infectious disease, and a rising rate of the isolation state of the date t;

the fitting state data comprises fitting susceptible state data, fitting exposed state data, fitting susceptible pathological state data and fitting isolated state data.

In the above technical solution, the state transition relationship of the infectious disease includes:

wherein, s (t) + e (t) + i (t) + q (t) ═ N, infection rate β ═ R_t(t) fitting susceptibility data on date t, (e) fitting exposure data on date t, (i) (t) fitting susceptibility data on date t, (q) (t) fitting isolation data on date t, and (R) isolation rate γ 1/TI_tEffective infection number representing date t, TE represents latency duration of the infectious disease, TI represents quarantine duration of the infectious disease, and the fitting status data includes the fitting susceptibility status data, the fitting exposure status data, the fitting susceptibility status data, and the fitting intervalAnd (4) off-state data.

In the above technical solution, the fitting case data includes fitting confirmed cases;

the extraction module is further used for extracting the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data which are in one-to-one correspondence with the dates from the fitting state data of the dates, and extracting the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data which are in one-to-one correspondence with the dates

And summing the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data which are in one-to-one correspondence with the plurality of dates, and taking the summed result as the fitting confirmed case in one-to-one correspondence with the plurality of dates.

In the above technical solution, the update module is further configured to construct an error function of the infectious disease trend prediction model according to a difference between real case data of the infectious disease on multiple dates and fitted case data of the multiple dates;

and updating the parameters of the infectious disease trend prediction model until the error function converges.

In the above technical solution, the parameters of the infectious disease trend prediction model include a basic infection number, a latency period duration, an isolation period duration of the infectious disease, an expected effective infection number at the end of the infectious disease, and a reduction rate of the effective infection number;

the updating module is further used for determining the updating directions of the basic infection number, the latent period duration, the isolation period duration, the effective infection number at the end of the infectious disease and the descending rate of the effective infection number in the infectious disease trend prediction model when the value of the error function is larger than the threshold value of the error function, and determining the updating directions of the basic infection number, the latent period duration, the effective infection number at the end of the infectious disease and the descending rate of the effective infection number in the infectious

And updating the basic infection number, the incubation period duration, the isolation period duration, the effective infection number at the end of the infectious disease and the reduction rate of the effective infection number according to the updating direction until the error function meets a convergence condition.

The embodiment of the invention provides an infectious disease epidemic situation trend prediction method based on an infectious disease trend prediction model, which comprises the following steps:

attenuating the basic infection number of the infectious disease according to the transmission time to obtain effective infection numbers corresponding to a plurality of prediction dates of the infectious disease;

determining state data which respectively correspond to the plurality of prediction dates and satisfy the constraint condition as fitting state data of the plurality of prediction dates by taking the state conversion relation included in the infectious disease trend prediction model as a constraint condition and taking the effective infection numbers of the plurality of prediction dates as known quantities in the constraint condition;

adding the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data in the fitting state data of the plurality of prediction dates to obtain the fitting confirmed cases of the plurality of prediction dates, and

and generating epidemic situation trends of the infectious diseases in the plurality of prediction dates according to the quantity change condition of the fitted diagnosed cases of the plurality of prediction dates.

The embodiment of the invention provides an infectious disease epidemic situation trend prediction device based on an infectious disease trend prediction model, which comprises:

a first determination module, configured to attenuate the basic infection count of the infectious disease according to a transmission time to obtain an effective infection count corresponding to a plurality of predicted dates of the infectious disease;

a second determination module, configured to determine, as fitting state data of the plurality of predicted dates, state data that respectively correspond to the plurality of predicted dates and satisfy the constraint condition, using the state conversion relationship included in the infectious disease trend prediction model as a constraint condition, and using the effective infection counts of the plurality of predicted dates as known quantities in the constraint condition;

and the processing module is used for summing the fitting exposed state data, the fitting infection state data and the fitting isolated state data in the fitting state data of the plurality of prediction dates to obtain fitting confirmed cases of the plurality of prediction dates, and generating epidemic situation trends formed by the infectious disease in the plurality of prediction dates according to the quantity change conditions of the fitting confirmed cases of the plurality of prediction dates.

In the above technical solution, the first determining module is further configured to add the expected effective infection counts at the end of the infectious disease and the attenuation results of the basic infection counts corresponding to the plurality of prediction dates to obtain the effective infection counts of the plurality of prediction dates, respectively;

the second determination module is further configured to substitute the effective infection counts for the plurality of predicted dates in the following formula of the state transition relationship to obtain fitted exposure state data, fitted infection state data, and fitted isolation state data for the plurality of predicted dates:

wherein s (t) + e (t) + i (t) + q (t) ═ N, β ═ R_t(t) fitting susceptibility data for prediction date t, e (t) fitting exposure data for prediction date t, i (t) fitting susceptibility data for prediction date t, q (t) fitting isolated state data for prediction date t, R (t) fitting isolated state data for prediction date t, and_tthe effective infection number of the prediction date t is shown, TE is the latent period duration of the infectious disease, and TI is the quarantine period duration of the infectious disease.

The embodiment of the invention provides a COVID-19 epidemic situation trend prediction method based on a 2019 coronavirus pneumonia COVID-19 model, which comprises the following steps:

determining effective infection counts corresponding to a plurality of predicted dates of the COVID-19 according to the characteristic that the effective infection counts of the COVID-19 decay with propagation time;

determining state data which respectively correspond to the plurality of prediction dates and satisfy the constraint condition as fitting state data of the plurality of prediction dates by using a state conversion relation included in the COVID-19 model as a constraint condition and using the effective infection numbers of the plurality of prediction dates as known quantities in the constraint condition;

determining inflection points of newly diagnosed cases of COVID-19 in the plurality of prediction dates based on changes in the number of fitted diagnosed cases for the plurality of prediction dates.

The embodiment of the invention provides a COVID-19 epidemic situation trend prediction device based on a COVID-19 model, which comprises:

a first determining module, configured to determine the effective infection counts corresponding to the plurality of predicted dates of the codv-19 according to a characteristic of the effective infection counts of the codv-19 decaying with propagation time;

a second determination module, configured to determine, as fitting state data of the plurality of prediction dates, state data that respectively correspond to the plurality of prediction dates and satisfy the constraint condition, using a state transition relationship included in the COVID-19 model as a constraint condition, and using the effective infection counts of the plurality of prediction dates as known quantities in the constraint condition;

and the processing module is used for summing the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data in the fitting state data of the plurality of prediction dates to obtain the fitting diagnosed cases of the plurality of prediction dates, and determining inflection points of newly added diagnosed cases of the COVID-19 in the plurality of prediction dates according to the quantity change condition of the fitting diagnosed cases of the plurality of prediction dates.

The embodiment of the invention provides electronic equipment for constructing an infectious disease trend prediction model, which comprises:

a memory for storing executable instructions;

and the processor is used for realizing the method for constructing the infectious disease trend prediction model provided by the embodiment of the invention when executing the executable instructions stored in the memory.

The embodiment of the invention provides electronic equipment for predicting epidemic situation trend of infectious diseases, which comprises:

a memory for storing executable instructions;

and the processor is used for realizing the method for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model provided by the embodiment of the invention when the executable instructions stored in the memory are executed.

The embodiment of the invention provides electronic equipment for predicting the trend of COVID-19 epidemic situations, which comprises:

a memory for storing executable instructions;

and the processor is used for implementing the COVID-19 epidemic situation trend prediction method based on the COVID-19 model provided by the embodiment of the invention when the processor executes the executable instructions stored in the memory.

Embodiments of the present invention provide a computer-readable storage medium storing executable instructions for causing a processor to implement a method for constructing an infectious disease trend prediction model according to embodiments of the present invention when the processor executes the executable instructions.

The embodiment of the invention provides a computer-readable storage medium, which stores executable instructions for causing a processor to execute the method for predicting the epidemic situation of an infectious disease based on the infectious disease trend prediction model provided by the embodiment of the invention.

The embodiment of the invention provides a computer-readable storage medium, which stores executable instructions and is used for causing a processor to execute the method for predicting the COVID-19 epidemic situation trend based on the COVID-19 model provided by the embodiment of the invention.

The embodiment of the invention has the following beneficial effects:

effective infection numbers of a plurality of dates decaying along with the propagation time are introduced into the state conversion relation of the infectious diseases included in the infectious disease trend prediction model to obtain fitting state data of the plurality of dates in one-to-one correspondence with the effective infection numbers of the plurality of dates, so that the effective infection numbers of the infectious diseases changing along with the time can be accurately simulated; furthermore, an accurate infectious disease trend prediction model can be constructed through the difference between a small amount of real case data and corresponding fitting case data, the epidemic situation trend can be accurately predicted at the epidemic situation initial stage of the infectious disease, and the method has an important reference value for epidemic situation prevention and control work.

Drawings

Fig. 1 is a schematic view of an application scenario of an infectious disease epidemic situation trend prediction system 10 according to an embodiment of the present invention;

FIG. 2 is a schematic structural diagram of an electronic device 500 for building an infectious disease trend prediction model according to an embodiment of the present invention;

FIGS. 3A-3B are schematic flow charts of methods for constructing an infectious disease trend prediction model according to embodiments of the present invention;

FIG. 4 is a schematic flowchart of an infectious disease epidemic situation trend prediction method based on an infectious disease trend prediction model according to an embodiment of the present invention;

fig. 5 is a schematic diagram illustrating the conversion between various states of the population in the D-SEIQ model according to an embodiment of the present invention;

FIG. 6 is a schematic flow chart of a method for parameter learning and epidemic situation trend prediction by the D-SEIQ model according to the embodiment of the present invention;

FIGS. 7A-11B are schematic diagrams illustrating the results of the verification of the optimal D-SEIQ model in the areas nationwide except for provinces A according to the present invention;

FIG. 12 is a schematic diagram of a fitted curve of fundamental propagation numbers provided by an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail with reference to the accompanying drawings, the described embodiments should not be construed as limiting the present invention, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

In the description that follows, references to the terms "first", "second", and the like, are intended only to distinguish similar objects and not to indicate a particular ordering for the objects, it being understood that "first", "second", and the like may be interchanged under certain circumstances or sequences of events to enable embodiments of the invention described herein to be practiced in other than the order illustrated or described herein.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to be limiting of the invention.

Before further detailed description of the embodiments of the present invention, terms and expressions mentioned in the embodiments of the present invention are explained, and the terms and expressions mentioned in the embodiments of the present invention are applied to the following explanations.

1) D-SEIQ model: a dynamic prediction model of infectious disease combines an improved infectious disease model and a machine learning parameter fitting technology, a D-SEIQ model divides a population into Susceptible persons (Susceptible state), exposed persons (exposed state), infected persons (Susceptible state) and isolated persons (isolated state), wherein the Susceptible persons (Susceptible), namely the persons who do not suffer from the disease but lack the immunity, are Susceptible to infection after being contacted with the Susceptible persons; exponents (exposeds), i.e. patients with latent stage, are suitable for infectious diseases with long latent stage; patients (infections), i.e., persons infected with Infectious disease, may be infected with a susceptible condition that changes it into an exposer or a patient; the isolater is the crowd who is in isolation during the epidemic situation, and the isolater can be the person that the sick person carries out the isolation through a period of time.

2) Basic replication Number (Basic replication Number): without external interference and without immunity in the population, an infected patient can be transmitted to a number of other individuals.

3) Effective infection Number (Effective replication Number): on the basis of the basic infection count, the infection count after the epidemic prevention measure, i.e., the number of individuals to whom an infected patient can be infected in the case of the epidemic prevention measure, is taken into consideration. Whether the epidemic situation is controlled depends on whether the effective infection number can be continuously less than 1, and if the effective infection number is continuously less than 1, the epidemic situation is controlled.

4)2019 coronavirus pneumonia (Corona Virus Disease 2019, COVID-19): the medicine is called new coronary pneumonia for short, and the world health organization is named as 2019 coronavirus disease, and the medicine refers to pneumonia (acute respiratory infectious disease) caused by 2019 novel coronavirus infection.

5) Inflection Point (Inflection Point): in the curve of the epidemic situation trend (the abscissa is the date, and the ordinate is the number of newly-added fitted confirmed cases), the corresponding tangent line crosses the point of the curve (the concave-convex dividing point of the curve). The inflection point is an important time node in the epidemic situation, for example, the inflection point can be a time point when the number of newly added fitting confirmed cases switches from an ascending trend to a descending trend, which indicates that the epidemic situation develops in a good direction; or adding a time point for switching the number of the fitted diagnosed cases from a descending trend to an ascending trend, which indicates that the epidemic situation trend develops in a serious direction.

The embodiment of the invention provides a method for constructing an infectious disease trend prediction model, an infectious disease epidemic situation trend prediction method based on the infectious disease trend prediction model, a device, electronic equipment and a computer readable storage medium, which can accurately determine fitting case data by combining effective infection numbers decaying along with propagation time.

An exemplary application of the electronic device for building an infectious disease trend prediction model provided by the embodiment of the invention is described below.

The electronic device for constructing the infectious disease trend prediction model provided by the embodiment of the invention can be various types of terminal devices or servers, taking a server as an example, the electronic device can be a server cluster deployed at the cloud end, and can open cloud services to medical staff, government health department staff or the public society, wherein a program for constructing the infectious disease trend prediction model and a program for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model are packaged. After medical staff, government health department staff or the public society inputs real case data of multiple dates of infectious diseases in open cloud service, a server deployed at the cloud calls a packaged program for constructing an infectious disease trend prediction model, parameters of the infectious disease trend prediction model are updated according to the difference between the real case data of the multiple dates and fitted case data of the multiple dates simulated by the infectious disease trend prediction model to realize the construction of the infectious disease trend prediction model, the packaged program for predicting the infectious disease epidemic situation trend based on the infectious disease trend prediction model is called, the fitted confirmed cases of the infectious diseases on the multiple prediction dates and the inflection points of newly-increased confirmed cases in the multiple prediction dates are predicted according to the built infectious disease trend prediction model, so that the government or the public can perform corresponding prevention and control measures according to the epidemic situation trend of the infectious diseases (such as the inflection points of the newly-increased confirmed cases), including organization/allocation of medical personnel, production/allocation of medical materials, organization and management of social production activities, and the like.

Referring to fig. 1, fig. 1 is a schematic view of an application scenario of an infectious disease epidemic situation trend prediction system 10 according to an embodiment of the present invention, a terminal 200 is connected to a server 100 through a network 300, and the network 300 may be a wide area network or a local area network, or a combination of both.

The terminal 200 may be used to acquire real case data for a plurality of dates, for example, a medical person inputs real case data for a plurality of dates through the terminal, and the terminal automatically acquires real case data for a plurality of dates after the input is completed.

In some embodiments, the terminal 200 locally performs the method for constructing the infectious disease trend prediction model and the method for predicting the infectious disease epidemic situation trend based on the infectious disease trend prediction model provided by the embodiments of the present invention to predict inflection points of newly-determined cases of infectious diseases according to the real case data of multiple dates of infectious diseases, for example, an infectious disease prediction Application (APP) is installed on the terminal 200, after the medical staff inputs the real case data of multiple dates of an infectious disease by the APP, the terminal 200 simulates the fitting case data of multiple dates in the state transformation relationship of infectious diseases included in the initial infectious disease trend prediction model, updates the parameters of the infectious disease trend prediction model according to the difference between the input real case data of multiple dates and the fitting case data of multiple dates simulated by the infectious disease trend prediction model, the construction of the infectious disease trend prediction model is realized, the fitted diagnosed cases of the infectious disease on a plurality of prediction dates and the inflection point dates in a plurality of prediction dates are predicted according to the constructed infectious disease trend prediction model, and the fitted diagnosed cases and the inflection point dates on the plurality of prediction dates are displayed on the display interface 210 of the terminal 200.

In some embodiments, the terminal 200 may also send, to the server 100 in the cloud via the network 300, real case data of multiple dates input by medical staff on the terminal 200, and invoke an infectious disease trend prediction function (an encapsulated program for constructing an infectious disease trend prediction model and an infectious disease trend prediction program based on the infectious disease trend prediction model) provided by the server 100, and the server 100 predicts inflection points of newly-added confirmed cases of an infectious disease according to the real case data of multiple dates of the infectious disease by using the method for constructing an infectious disease trend prediction model and the infectious disease trend prediction method based on the infectious disease trend prediction model provided by the embodiment of the present invention, for example, an infectious disease prediction application is installed on the terminal 200, and the medical staff inputs the real case data of multiple dates of a certain infectious disease in the infectious disease prediction application, the terminal 200 sends the real case data of the infectious disease on a plurality of dates to the server 100 through the network 300, after the server 100 receives the real case data of the infectious disease on a plurality of dates, the server 100 calls the packaged program for constructing the infectious disease trend prediction model, updates the parameters of the infectious disease trend prediction model according to the difference between the input real case data of the plurality of dates and the fitted case data of the plurality of dates simulated by the infectious disease trend prediction model so as to complete the construction of the infectious disease trend prediction model, calls the packaged program for predicting the infectious disease epidemic situation trend of the infectious disease based on the infectious disease trend prediction model, predicts the fitted confirmed cases of the plurality of prediction dates and the inflection point date of the newly-added confirmed case in the plurality of prediction dates according to the constructed infectious disease trend prediction model, and returns the fitted confirmed cases and the inflection point date of the plurality of prediction dates to the infectious disease prediction application, and displaying the fitted diagnosed cases and inflection points of the plurality of predicted dates on the display interface 210 of the terminal 200, or the server 100 directly gives the fitted diagnosed cases and inflection points of the plurality of predicted dates.

The following describes a structure of an electronic device for constructing an infectious disease trend prediction model according to an embodiment of the present invention, where the electronic device for constructing the infectious disease trend prediction model may be various terminals, such as a mobile phone, a computer, and the like, and may also be the server 100 shown in fig. 1.

Referring to fig. 2, fig. 2 is a schematic structural diagram of an electronic device 500 for building an infectious disease trend prediction model according to an embodiment of the present invention, and taking the electronic device 500 as a server as an example, the electronic device 500 for building an infectious disease trend prediction model shown in fig. 2 includes: at least one processor 510, memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together by a bus system 540. It is understood that the bus system 540 is used to enable communications among the components. The bus system 540 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, however, the various buses are labeled as bus system 540 in fig. 2.

The Processor 510 may be an integrated circuit chip having Signal processing capabilities, such as a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like, wherein the general purpose Processor may be a microprocessor or any conventional Processor, or the like.

The memory 550 may comprise volatile memory or nonvolatile memory, and may also comprise both volatile and nonvolatile memory. The non-volatile Memory may be a Read Only Memory (ROM), and the volatile Memory may be a Random Access Memory (RAM). The memory 550 described in connection with embodiments of the invention is intended to comprise any suitable type of memory. Memory 550 optionally includes one or more storage devices physically located remote from processor 510.

In some embodiments, memory 550 can store data to support various operations, examples of which include programs, modules, and data structures, or subsets or supersets thereof, as exemplified below.

An operating system 551 including system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks;

a network communication module 552 for communicating to other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: bluetooth, wireless compatibility authentication (WiFi), and Universal Serial Bus (USB), etc.;

in some embodiments, the apparatus for constructing an infectious disease trend prediction model provided by the embodiments of the present invention may be implemented in software, and fig. 2 shows an apparatus 555 for constructing an infectious disease trend prediction model stored in a memory 550, which may be software in the form of programs and plug-ins, and includes a series of modules including an attenuation module 5551, a determination module 5552, an extraction module 5553, and an update module 5554; the attenuation module 5551, the determination module 5552, the extraction module 5553 and the update module 5554 are used for implementing the function of constructing the infectious disease trend prediction model provided by the embodiment of the invention.

As can be understood from the foregoing, the method for building an infectious disease trend prediction model according to the embodiments of the present invention may be implemented by various types of electronic devices, such as an intelligent terminal, a server, and the like, for building an infectious disease trend prediction model.

The method for constructing the infectious disease trend prediction model provided by the embodiment of the invention is described below by combining the exemplary application and implementation of the server provided by the embodiment of the invention. Referring to fig. 3A, fig. 3A is a schematic flowchart of a method for constructing an infectious disease trend prediction model according to an embodiment of the present invention, which is described with reference to the steps shown in fig. 3A.

In step 101, the basic infection count of the infectious disease is attenuated according to the transmission time, and effective infection counts on a plurality of dates during the transmission period are obtained.

As an example of acquiring real medical record data, a medical staff may input real case data of multiple dates of an infectious disease, for example, real diagnosed cases of 2019-01-20 to 2019-01-26, on an input interface of a terminal, and after the input is completed, the terminal may forward the real case data of multiple dates to a server, so that the server constructs an infectious disease trend prediction model according to the real case data of multiple dates.

As a pre-processing link for constructing an infectious disease trend prediction model, the basic infectious number of the infectious disease can be attenuated according to the transmission time so as to obtain the effective infectious number corresponding to a plurality of dates. Wherein the date may be continuous or intermittent.

In the present example, it was found that the effective infection number of infectious diseases has a characteristic of being decayed with time. In order to simulate the characteristics of the infectious diseases in the infectious disease trend prediction model, the basic infectious number of the infectious diseases is attenuated according to the propagation time, so that the effective infectious number attenuated along with time is obtained, the effective infectious number is closer to the real infectious disease situation, and the fitting case data can be accurately determined to construct the infectious disease trend prediction model.

Referring to fig. 3B, fig. 3B is an alternative flowchart of a method for constructing an infectious disease trend prediction model according to an embodiment of the present invention, and fig. 3B shows that step 101 in fig. 3A can be implemented by step 1011 shown in fig. 3B. In step 1011, the expected effective infection counts at the end of the infection and the attenuation results corresponding to the basic infection counts on a plurality of dates are added to obtain effective infection counts on a plurality of dates; wherein the attenuation result is obtained by attenuating the basic infection number based on the transmission time of the infectious disease.

For example, after determining the attenuation result corresponding to the basic infection number in a certain day, the expected effective infection number at the end of the infection and the attenuation result corresponding to the basic infection number in a certain day are added to obtain the effective infection number of the day, for example, the formula of the effective infection number of the day is R_t＝R_∞+R₀*e^-tWherein R is_tEffective number of infections, R, representing date t₀*e^-tResults of decay of the basal infection number on the date t, R₀Denotes the number of basic infections, R_∞Indicating the number of effective infections expected at the end of the infection. Wherein, with respect to R_t＝R_∞+R₀*e^-tAre applicable to embodiments of the present invention.

In the examples of the present invention, it was found that the effective infection number of some infectious diseases exponentially decreased with time. To simulate such a characteristic of an infectious disease in the infectious disease trend prediction model, in some embodiments, the expected effective infection count at the end of the infectious disease and the decay results of the basic infection count corresponding to a plurality of dates are added to obtain effective infection counts for the plurality of dates, respectively, including: according to the formula R_t＝R_∞+R₀*e^-θ*tDetermining the effective infection number of any one of a plurality of dates t, wherein R_tEffective number of infections, R, representing date t₀*e^-θ*tResults of decay of the basal infection number on the date t, R₀Representing the basic infection number, theta representing the rate of decrease of the effective infection number, R_∞Indicating the number of effective infections expected at the end of the infection.

In connection with the above example, to better simulate a real effective infection count, a dynamic parameter θ is introduced into the effective infection count such that R_tCan exponentially decline with time to fit the trend of exponentially decline in effective infection numbers in real infectious diseases with time.

In step 102, fitting state data of a plurality of dates in one-to-one correspondence with effective infection numbers of a plurality of dates is determined in the state transition relation of the infectious disease included in the infectious disease trend prediction model.

The infectious disease trend prediction model is used for predicting epidemic situation trends of infectious diseases. After the server determines the effective infection counts on a plurality of dates, the effective infection counts on the plurality of dates can be introduced into the state conversion relation of the infectious diseases included in the infectious disease trend prediction model to determine the fitting state data of the plurality of dates corresponding to the effective infection counts on the plurality of dates one by one, so that the fitting case data of the plurality of dates can be obtained according to the fitting state data of the plurality of dates. Wherein the infectious disease state can include a susceptible state, an exposed state, an infected state and an isolated state; the infectious disease state may also include a susceptible state, an exposed state, an infected state, a rehabilitated state, and the like.

Referring to fig. 3B, fig. 3B is an alternative flowchart of a method for constructing an infectious disease trend prediction model according to an embodiment of the present invention, and fig. 3B illustrates that step 101 in fig. 3A can be implemented by step 1021 shown in fig. 3B. At step 1021, using the state transition relationship of the infectious disease as a constraint and the effective infectious counts on a plurality of dates as a known amount in the constraint; in step 1022, status data which correspond to a plurality of dates one by one and satisfy the constraint condition is determined as fitting status data of the plurality of dates; the state conversion relation of the infectious diseases comprises parameters of an infectious disease trend prediction model.

As an example, the state transition relationship of infectious diseases is used as a constraint condition, that is, a functional relationship, effective infectious numbers on a plurality of dates are used as known quantities in the functional relationship, unknown quantities, that is, state data which are in one-to-one correspondence with the plurality of dates and satisfy the functional relationship are obtained from the functional relationship and the known quantities, and the state data are used as fitting state data of the plurality of dates. The state conversion relation of the infectious diseases comprises parameters of an infectious disease trend prediction model.

In some embodiments, the infectious disease trend prediction model includes state transition relationships including: the conversion relation among the susceptible state of the date t, the infected state of the date t, the infection speed of the infectious disease and the descending rate of the susceptible state of the date t, the infected state of the date t, the exposed state of the date t, the infection speed of the infectious disease, the incidence speed of the infectious disease and the ascending rate of the exposed state of the date t, the conversion relation among the exposed state of the date t, the infected state of the date t, the incidence speed of the infectious disease, the isolation speed of the infectious disease and the ascending rate of the infected state of the date t, and the conversion relation among the infected state of the date t, the isolation speed of the infectious disease and the ascending rate of the isolated state of the date t; the fitting state data comprises fitting susceptible state data, fitting exposed state data, fitting susceptible diseased state data and fitting isolated state data.

For example, by substituting the effective infection number of the date t into the state conversion relationship, the susceptible state of the date t, the exposed state of the date t, the infected state of the date t, and the isolated state of the date t can be obtained, and the susceptible state of the date t is used as the fitting susceptible state data of the date t, the exposed state of the date t is used as the fitting exposed state data of the date t, the susceptible state of the date t is used as the fitting susceptible state data of the date t, and the isolated state of the date t is used as the fitting isolated state data of the date t.

In order to accurately simulate the conversion of four states in real infectious diseases, infectious disease transmission factors such as infection speed, disease incidence speed and the like are introduced into the state conversion relation of the infectious diseases.

As an example, in some embodiments, the state transition relationship of infectious disease includes:

wherein, s (t) + e (t) + i (t) + q (t) ═ N, infection rate β ═ R_t(t) fitting susceptibility data on date t, (e) fitting exposure data on date t, (i) (t) fitting susceptibility data on date t, (q) (t) fitting isolation data on date t, and (R) isolation rate γ 1/TI_tThe effective infection number of the date t is shown, TE is shown as the latent period duration of the infectious disease, TI is shown as the isolation period duration of the infectious disease, and the fitting state data comprises fitting susceptibility state data, fitting exposure state data, fitting susceptibility state data and fitting isolation state data.

In the embodiment of the invention, the model provided by the related technology cannot well fit the reality of the infectious disease, and particularly, the influence degrees of different populations are different in the epidemic situation of the infectious disease. In order to fit the reality of the infectious disease in the infectious disease trend prediction model, the embodiment of the invention specifically provides four states in the infectious disease prediction model, including a susceptible state, an exposed state, an infected state and an isolated state, and the degrees of the influence of the susceptible state, the exposed state, the infected state and the isolated state on the population are sequentially increased. Therefore, the influence of the infectious diseases on different crowds can be accurately and quantitatively reflected.

In addition, the conversion relation of the four states in the real infectious disease is simulated through the mathematical relational expression, and the data of the four states of the infectious disease is obtained through a method of carrying out equation solving on the mathematical relational expression, so that on one hand, the problem of unexplainable property of a processing process (because the machine learning model is used for predicting the four states of the infectious disease, the credibility of a prediction result can be questioned) when a pure machine learning model is used for predicting the four states of the infectious disease can be avoided, and the credibility of the prediction result is ensured; on the other hand, the influence degree of infectious diseases in the real environment on different populations is simulated through the susceptible state, the exposed state, the infected state and the isolated state, so that the accuracy of predicting confirmed cases is ensured; on the other hand, compared with a machine learning model mode, the mathematical equation solving mode for predicting the confirmed cases obviously reduces the calculation scale, and can save the calculation resources of the server, so that a large amount of data collected during the epidemic situation can be quickly responded, the confirmed cases can be predicted in time, and effective reference is provided for epidemic situation prevention and control work.

In step 103, fitting case data on a plurality of dates is extracted from the fitting state data on a plurality of dates.

The fitting state data comprises fitting susceptible state data, fitting exposed state data, fitting susceptible pathological state data and fitting isolated state data. The case data may be confirmed case data, isolated case data, susceptible case data, or the like. The fitting state data and the fitting case data are related to each other, and fitting case data on a plurality of dates can be extracted from the fitting state data on the plurality of dates.

In the embodiment of the invention, compared with the related technology in which the learning mode of the confirmed case is simulated and the mode of fitting the confirmed case is directly obtained, the mode of fitting the confirmed case is obtained by adding the fitting exposure state data, the fitting susceptibility state data and the fitting isolation state data in the fitting state data, and as the fitting exposure state data, the fitting susceptibility state data and the fitting isolation state data can truly simulate the state of the infectious disease, the fitting confirmed case can be accurately obtained through a simple mathematical mode, so that the trend of the infectious disease can be accurately predicted in the subsequent process.

In some embodiments, fitting the case data comprises fitting a diagnosed case; extracting fitting case data of a plurality of dates from the fitting state data of the plurality of dates, and the method comprises the following steps: extracting the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data which are in one-to-one correspondence with the dates from the fitting state data of the dates, adding the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data which are in one-to-one correspondence with the dates, and taking the addition result as the fitting diagnosed case which is in one-to-one correspondence with the dates.

As an example, the fitting exposed state data of the date t, the fitting pathological state data of the date t and the fitting isolated state data of the date t are extracted from the fitting state data of the date t, and the sum of the fitting exposed state data of the date t, the fitting pathological state data of the date t and the fitting isolated state data of the date t is used as a fitting confirmed case of the date t, namely the fitting exposed state data of the date t + the fitting pathological state data of the date t + the fitting isolated state data of the date t is the fitting confirmed case of the date t.

In step 104, parameters of the infectious disease trend prediction model are updated according to the difference between the real case data of the plurality of dates and the fitting case data of the plurality of dates, and the updated parameters are used as parameters for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model.

For example, after the server obtains the fitted case data of a plurality of dates, the server determines the difference between the received real case data of a plurality of dates and the fitted case data of a plurality of dates, updates the parameters of the infectious disease trend prediction model according to the difference, and takes the updated parameters as the parameters used for predicting the epidemic situation trend of the infectious disease based on the infectious disease trend prediction model. Updating parameters of the infectious disease trend prediction model can be realized through a small amount of real case data and corresponding fitting case data so as to complete the construction of the infectious disease trend prediction model, and the infectious disease epidemic situation trend is predicted through the constructed infectious disease trend prediction model.

In some embodiments, updating parameters of the infectious disease trend prediction model based on differences between real case data for multiple dates of infectious disease and fitted case data for multiple dates comprises: constructing an error function of an infectious disease trend prediction model according to the difference between real case data of multiple dates of infectious diseases and fitting case data of multiple dates; and updating the parameters of the infectious disease trend prediction model until the error function converges.

Wherein, after the server obtains the fitting case data of a plurality of dates, the real case data of a plurality of dates and the fitting case data of a plurality of dates are usedAnd determining the difference between the parameters, namely determining the value of an error function of the infectious disease trend prediction model, judging whether the value of the error function exceeds a preset threshold value, determining an error signal of the infectious disease trend prediction model based on the error function when the value of the error function exceeds the preset threshold value, reversely transmitting error information in the infectious disease trend prediction model, and updating model parameters of each layer in the transmission process. Wherein the error function is

Wherein f is_iRepresenting fitted case data, y_iRepresenting the real case data and N representing the total date of the real case data or the fitted case data.

Describing backward propagation, inputting training sample data into an input layer of a neural network model, passing through a hidden layer, finally reaching an output layer and outputting a result, which is a forward propagation process of the neural network model, wherein because the output result of the neural network model has an error with an actual result, an error between the output result and the actual value is calculated and is propagated backward from the output layer to the hidden layer until the error is propagated to the input layer, and in the process of backward propagation, the value of a model parameter is adjusted according to the error; and continuously iterating the process until convergence, wherein the infectious disease trend prediction model belongs to a neural network model.

In some embodiments, the parameters of the infectious disease trend prediction model include a base infection count of the infectious disease, a latency period duration, an isolation period duration, a desired effective infection count at the end of the infectious disease, and a rate of decline of the effective infection count; updating parameters of the infectious disease trend prediction model until the error function converges, including: when the value of the error function is larger than the threshold value of the error function, determining the updating direction of the basic infection number, the latent period duration, the isolation period duration, the effective infection number at the end of the infectious disease and the reduction rate of the effective infection number in the infectious disease trend prediction model, and updating the basic infection number, the latent period duration, the isolation period duration, the effective infection number at the end of the infectious disease and the reduction rate of the effective infection number according to the updating direction until the error function meets the convergence condition.

As an example, when the value of the error function is greater than the error function threshold, the update direction of the basic infection number, the latency period duration, the isolation period duration, the effective infection number at the end of the infectious disease, and the decrease rate of the effective infection number in the infectious disease trend prediction model is determined by a heuristic search method, so that the basic infection number, the latency period duration, the isolation period duration, the effective infection number at the end of the infectious disease, and the decrease rate of the effective infection number are updated according to the update direction until the error function is greater than or equal to the error function threshold. Wherein, the four combined parts for updating the infectious disease trend prediction model by the heuristic search method are respectively: 1) an objective function:

2) domain space: a parameter set consisting of 5 parameters (basic infection number, latency period duration, isolation period duration, effective infection number at the end of infection and reduction rate of effective infection number) to be fitted; 3) and (3) an optimization algorithm: a method of constructing a proxy model and selecting a value of a hyper-parameter to be evaluated next, i.e., a Tree Pazen Evaluator (TPE); 4) history data: the objective function evaluation results (5 parameters and validation loss) are stored.

The parameters of the infectious disease trend prediction model are solved in a machine learning model mode, and compared with the method that the parameters of the infectious disease trend prediction model in the related technology depend on experience setting, the method has better precision and depends on the strong computing power of the server, and the method can carry out real-time optimization on the infectious disease trend prediction model according to the updated basic infectious number of the infectious disease when the epidemic situation changes, so that the prediction result of the diagnosed case is updated in time according to the change of the epidemic situation, and good real-time performance is realized.

The infectious disease epidemic situation trend prediction method based on the infectious disease trend prediction model provided by the embodiment of the invention is described below by combining with the exemplary application and implementation of the server provided by the embodiment of the invention. The server for implementing the method for constructing the infectious disease trend prediction model and the server for implementing the infectious disease epidemic situation trend prediction method based on the infectious disease trend prediction model can be mutually independent electronic devices, namely the two mutually independent servers respectively and independently realize the function of constructing the infectious disease trend prediction model and the infectious disease epidemic situation trend prediction function; the server for implementing the method for constructing the infectious disease trend prediction model and the server for implementing the method for predicting the infectious disease epidemic situation trend based on the infectious disease trend prediction model may also be the same electronic device (i.e., the electronic device 500 integrates a program for constructing the infectious disease trend prediction model and a program for predicting the infectious disease epidemic situation trend based on the infectious disease trend prediction model, and after the infectious disease trend prediction model is constructed, the infectious disease epidemic situation trend is directly predicted according to the constructed infectious disease trend prediction model).

Taking the example of the electronic device 500 integrating the program for constructing the infectious disease trend prediction model and the program for predicting the infectious disease epidemic situation trend based on the infectious disease trend prediction model as shown in fig. 2, the infectious disease epidemic situation trend prediction apparatus 556 based on the infectious disease trend prediction model stored in the memory 550 includes a first determination module 5561, a second determination module 5562 and a processing module 5563; the first determining module 5561, the second determining module 5562 and the processing module 5563 are used for implementing the infectious disease epidemic situation trend prediction method based on the infectious disease trend prediction model provided by the embodiment of the invention.

The infectious disease epidemic situation trend prediction method based on the infectious disease trend prediction model provided by the embodiment of the invention is described below by combining with the exemplary application and implementation of the server provided by the embodiment of the invention. Referring to fig. 4, fig. 4 is a flowchart of an infectious disease epidemic situation trend prediction method based on an infectious disease trend prediction model according to an embodiment of the present invention, which is described with reference to the steps shown in fig. 4.

In step 201, the basic infection count of the infectious disease is attenuated according to the propagation time, and effective infection counts corresponding to a plurality of predicted dates of the infectious disease are obtained.

After the server obtains the constructed infectious disease trend prediction model, the basic infectious disease number of the infectious disease is attenuated according to the transmission time, so as to obtain the effective infectious disease number of a plurality of prediction dates (namely future dates, for example, the current date is 2020-01-20, and the prediction dates can be 2020-01-21, 2020-01-22, 2020-01-23 and the like) corresponding to the infectious disease. The forecast date may be continuous or discontinuous. The basic infection number of the infectious disease is attenuated according to the propagation time, so that the effective infection number attenuated along with the time is obtained, the effective infection number is closer to the real epidemic situation of the infectious disease, and the fitting case data can be accurately determined in the following process.

In some embodiments, it is found that the effective infection count of some infectious diseases is exponentially decreased with time, and the effective infection count of some infectious diseases is not completely equal to 0 after the epidemic situation is over. In order to make the effective infection number closer to the real epidemic situation of the infectious disease, the fitting case data can be accurately determined in the following.

Thus, in some embodiments, attenuating the base infection count of the infectious disease based on the time of transmission to obtain an effective infection count corresponding to a plurality of predicted dates of the infectious disease, comprises: adding the expected effective infection number at the end of the infectious disease and the attenuation results corresponding to the basic infection number on a plurality of prediction dates to obtain effective infection numbers on the plurality of prediction dates; wherein the attenuation result is obtained by attenuating the basic infection number based on the transmission time of the infectious disease.

For example, after determining the attenuation result corresponding to the basic infection number on a certain prediction date, the expected effective infection number at the end of the infection and the attenuation result corresponding to the basic infection number on a certain prediction date are added to obtain the effective infection number on the prediction date, for example, the formula of the effective infection number on the prediction date is R_t＝R_∞+R₀*e^-tWherein R is_tEffective infection number R representing predicted date t₀*e^-tResults of decay of the basal infection number at the predicted date t, R₀Denotes the number of basic infections, R_∞Indicating the number of effective infections expected at the end of the infection.

In step 202, the state transition relation included in the infectious disease trend prediction model is used as a constraint condition, the effective infection numbers of the plurality of prediction dates are used as known quantities in the constraint condition, and the state data which respectively correspond to the plurality of prediction dates and satisfy the constraint condition is determined as fitting state data of the plurality of prediction dates.

As an example, after the server determines the effective infection counts on the plurality of prediction dates, the effective infection counts on the plurality of prediction dates may be introduced into the state transition relationship of the infectious disease included in the infectious disease trend prediction model to determine fitting state data of the plurality of prediction dates in one-to-one correspondence with the effective infection counts on the plurality of prediction dates, so as to obtain fitting case data of the plurality of prediction dates from the fitting state data of the plurality of prediction dates.

In the embodiment of the present invention, it is found that the infectious disease has different spread degrees to people during epidemic situations, namely, four states exist, and the states of the infectious disease comprise a susceptible state, an exposed state, an infected state and an isolated state. The conversion relation of the four states in the real infectious diseases is simulated through the mathematical relational expression, the four states in the infectious diseases are obtained, the four states in the infectious diseases are prevented from being predicted by adopting a machine learning method, and the unexplainable property of machine learning is solved. And the real infectious diseases are simulated through the susceptible state, the exposed state, the susceptible state and the isolated state, accurate fitting state data can be obtained, and accurate fitting case data are obtained according to the fitting state data.

Therefore, in some embodiments, determining, as fitting state data of a plurality of prediction dates, state data that respectively correspond to the plurality of prediction dates and satisfy the constraint condition, with the state transition relationship included in the infectious disease tendency prediction model as the constraint condition and the effective infectious counts of the plurality of prediction dates as known quantities in the constraint condition, includes: substituting effective infection numbers of a plurality of prediction dates into the following formula of state conversion relation to obtain fitted exposure state data, fitted infection state data and fitted isolation state data of the plurality of prediction dates:

wherein s (t) + e (t) + i (t) + q (t) ═ N, β ═ R_t(t) fitting susceptibility data for prediction date t, e (t) fitting exposure data for prediction date t, i (t) fitting susceptibility data for prediction date t, q (t) fitting isolated state data for prediction date t, R (t) fitting isolated state data for prediction date t, and_tthe effective infection number at the predicted date t is shown, TE shows the latent period of the infectious disease, and TI shows the quarantine period of the infectious disease.

In step 203, the fitted exposed state data, the fitted sensitive state data and the fitted isolated state data in the fitted state data of the plurality of prediction dates are summed to obtain fitted confirmed cases of the plurality of prediction dates.

As an example, the fitted exposed state data of the prediction date t, the fitted diseased state data of the prediction date t and the fitted isolated state data of the prediction date t are extracted from the fitted state data of the prediction date t, and the sum of the fitted exposed state data of the prediction date t, the fitted diseased state data of the prediction date t and the fitted isolated state data of the prediction date t is used as a fitted diagnosed case of the prediction date t, namely the fitted exposed state data of the prediction date t, the fitted diseased state data of the prediction date t and the fitted isolated state data of the prediction date t are the fitted diagnosed case of the prediction date t.

In step 204, based on the number of cases to be diagnosed that are fitted to the plurality of prediction dates, an epidemic situation trend of the infectious disease is generated in the plurality of prediction dates.

As an example, the change of the number of the fitted confirmed cases on a plurality of prediction dates is analyzed to determine new confirmed cases each day, and the date corresponding to the peak value of the new confirmed cases is used as the inflection point of the new confirmed cases of the infectious disease, so that corresponding infectious disease epidemic prevention measures can be carried out according to the inflection point of the new confirmed cases of the infectious disease. For example, if the cases to be diagnosed by fitting at the plurality of prediction dates are 100 cases to be diagnosed by fitting (02-01), 150 cases to be diagnosed by fitting (02-02), 220 cases to be diagnosed by fitting (02-03), 300 cases to be diagnosed by fitting (02-04), 360 cases to be diagnosed by fitting (02-05) and 400 cases to be diagnosed by fitting (02-06), 50 cases to be diagnosed by newly-increased number 02-02, 70 cases to be diagnosed by newly-increased number 02-03, 80 cases to be diagnosed by newly-increased number 02-04, 60 cases to be diagnosed by newly-increased number 02-05 and 40 cases to be diagnosed by newly-increased number 02-06, 70 cases to be diagnosed by newly-increased number 02-03 are the peak, and 70-03 are the inflection points of newly-determined cases of infectious diseases.

For example, when the infection is COVID-19, and a COVID-19 model is constructed, the COVID-19 model is updated using the actual case data of COVID-19, that is, the basic infection number of COVID-19 is attenuated according to the transmission time, and the effective infection numbers of a plurality of dates during the transmission period are obtained; determining fitting state data of a plurality of dates corresponding to effective infection numbers of the dates in a state conversion relation of the COVID-19 included in the COVID-19 model; extracting fitting case data of a plurality of dates from the fitting state data of the plurality of dates; and updating parameters of the COVID-19 model according to the difference between the actual COVID-19 case data of a plurality of dates and the fitting case data of a plurality of dates, and using the updated parameters as parameters for predicting the COVID-19 epidemic trend based on the COVID-19 model. After the COVID-19 model is constructed, determining the effective infection numbers of a plurality of prediction dates corresponding to the COVID-19 according to the characteristic that the effective infection numbers of the COVID-19 decay along with the propagation time; determining state data which respectively correspond to the plurality of prediction dates and satisfy the constraint conditions as fitting state data of the plurality of prediction dates by taking the state conversion relation included in the COVID-19 model as a constraint condition and taking the effective infection numbers of the plurality of prediction dates as known quantities in the constraint condition; and adding the fitting exposed state data, the fitting susceptible state data and the fitting isolated state data in the fitting state data of the plurality of prediction dates to obtain fitting diagnosed cases of the plurality of prediction dates, and determining inflection points of newly-added diagnosed cases of the COVID-19 in the plurality of prediction dates according to the quantity change condition of the fitting diagnosed cases of the plurality of prediction dates.

In the following, an exemplary application of the embodiments of the present invention in a practical application scenario will be described.

The embodiment of the invention can be applied to application scenes of prediction of COVID-19 epidemic situation trend, as shown in figure 1, a terminal 200 is connected with a server 100 deployed at the cloud through a network 300, the COVID-19 prediction application is installed on the terminal 200, medical staff inputs real case data of a certain infectious disease on a plurality of dates in the infectious disease prediction application, the terminal 200 sends the real diagnosis case data of the COVID-19 to the server 100 through the network 300, after the server 100 receives the real case data of the COVID-19, in the state conversion relation of the C OVID-19, fitting state data (fitting exposure state data, fitting infection state data and fitting isolation state data) are determined, the fitting diagnosis case data are extracted from the fitting state data, and according to the difference between the input real diagnosis case data and the diagnosis case data, updating parameters of the COVID-19 model to obtain an optimal COVID-19 model, predicting the COVID-19 according to the optimal COVID-19 model to determine inflection points of the COVID-19 in future fitted diagnosed cases and newly diagnosed cases, returning the inflection points of the fitted diagnosed cases and newly diagnosed cases in future to the COV ID-19 prediction application, and displaying the inflection points of the fitted diagnosed cases and newly diagnosed cases in future on a display interface 210 of the terminal 200, so that a government or the public can perform corresponding COVID-19 epidemic prevention measures according to the inflection points of the predicted fitted diagnosed cases and newly diagnosed cases.

In the related technology, the prediction method of the epidemic situation trend of COVID-19 is mainly divided into two types: 1) direct use of models of infectious diseases; 2) an optimization-based machine learning method.

The method is characterized in that the COVID-19 epidemic situation is predicted to serve as a regression task in machine learning based on an optimized machine learning method, and the infection trend of the COVID-19 is predicted by training a Long Short-Term Memory (LSTM) network. The data of COVID-19 and the data of SARS-CoV in 2003 are combined together based on an optimized machine learning model to train the LSTM network, and finally the infection trend of COVID-19 is predicted according to the trained LSTM network. For example, 3-day newly infected cases of COVID-19 are input into the optimized machine learning model, the input dimension is (3, 1), the input dimension is transmitted into the LSTM unit of the hidden layer, and after the LSTM processes the 3-day newly infected time-series case data, the newly infected time-series case data is regressed by using a full connection layer. In addition, the optimized based machine learning model requires training through SARS-CoV data and then using the trained optimized based machine learning model to predict on COVID-19 data, resulting in a long-term infection case estimate of COVID-19 over the next 80 days.

Although, both infectious disease models and optimized machine learning based models enable COVID-19 epidemic prediction. However, the infectious disease model and the machine learning model based on optimization have the following problems: 1) the infectious disease model is sensitive to parameters and the accuracy of the infectious disease model is highly dependent on an accurate estimate of the transmission characteristics (e.g., effective infection count, latency and infection period, etc.) of COVID-19. In reality, due to the interference of a plurality of external forces (such as government regulation, virus variation and the like), the factors are difficult to accurately estimate, and the difficulty of infectious disease model prediction is increased; 2) the exposer in the infectious disease model has no infectivity, but the COVID-19 has infectivity in the latent period, so the infectious disease model cannot accurately simulate the infection process of the COVID-19; 3) the amount of data and feature quantity available for COVID-19 are relatively small, but the optimization-based machine learning method relies heavily on data, so that the optimization-based machine learning method easily generates overfitting (over fitting) on the trend prediction of COVID-19, the prediction is inaccurate, and a model trained completely based on the optimization-based machine learning method is lack of interpretability.

In order to solve the problems, the invention provides a COVID-19 infectious disease dynamic prediction model D-SEIQ (infectious disease trend prediction model) combining an improved infectious disease model and a machine learning parameter fitting technology, wherein the D-SEIQ model comprises a susceptible state, an exposed state, a diseased state and an isolated state, the real infectious disease condition can be better fitted through the isolated state, in addition, various interventions in the real infectious disease can be reflected by introducing dynamic parameters into the effective infectious number, and the D-SEIQ model can accurately predict the development trend of the COVID-19 epidemic situation. Among them, Susceptible (susceptable), i.e., a population who is not suffering from a disease but lacks immune competence and is Susceptible to infection after being contacted with a disease-Susceptible person; exposure (expesed), i.e. latent patients, is suitable for infectious diseases with long latent periods; an Infectious disease (infection), i.e. a person infected with an Infectious disease, can infect a susceptible state to change it into an exposed state or an Infectious state; isolated state, i.e. isolated population.

The embodiment of the invention predicts the long-term development trend of the COVID-19 epidemic situation through the dynamic interpretable D-SEIQ model and has the following significance: 1) the anxiety of the public can be properly relieved, the large-scale epidemic diseases influence the lives of most people, and the prediction of the epidemic situation trend can help the public to know the development of the epidemic situation and reasonably arrange the life; 2) the method can help the government to evaluate the effectiveness of the policy, and for different control measures possibly issued by governments in various regions of epidemic situations, the parameters learned in the D-SEIQ model can help the government to know whether the measures are effective or not and guide the government to properly adjust the policy; 3) can help public health safety experts to know the characteristics of the virus, and parameters such as effective infection number in a D-S EIQ model can provide more virus information about COVID-19 for related experts. The implementation of the D-SEIQ model is described in detail below:

the D-SEIQ model comprises a susceptible state, an exposed state, a diseased state and an isolated state, the real infectious disease condition can be better fitted through the isolated state, and in addition, dynamic parameters are introduced into the effective infectious number, so that the effective infectious number can be gradually reduced along with time. Because asymptomatic infectors exist in the COVID-19 epidemic situation, and a medical institution lacks a targeted treatment scheme at the initial stage of the epidemic situation, and the government adopts extremely strong and effective control in epidemic situation prevention and control, the D-SEIQ model comprises an isolation state Q (quantified), the infectious disease dynamic diagram of the D-SEIQ model is shown in figure 5, the D-SEIQ model comprises a susceptible state S, an exposed state E, an susceptible state I and an isolation state Q, and the susceptible state S can be converted into the exposed state E after contacting with a type I infected patient or the exposed state E; the exposed state E can be converted into a diseased state I after a period of incubation; after a period of time, the diseased state I can be converted to the isolated state Q. Wherein, the differential equation of the conversion between each state of the crowd in the D-SEIQ model is shown in the formulas (5) to (8):

wherein s (t) + e (t) + i (t) + q (t) ═ N, β denotes the infection rate, β ═ R_t(t) sensitive state data indicating date t, (e) exposed state data indicating date t, (i) sensitive state data indicating date t, (q), (t) isolated state data indicating date t, and (R) isolated state data indicating date t_tThe number of effective infections on date t, TE the duration of incubation of COVID-19, and TI the duration of the quarantine period of COVID-19. The parameter gamma in the D-SEIQ model represents the isolation speed, and can reflect some external factors, such as medical resources, detection reagents, government control strength and the like, when the medical resources are abundant, the detection reagent detection efficiency is high, and the government control strength is strong, the isolation speed is correspondingly high, and the parameter gamma can help. The D-SEIQ model better models the propagation of COVID-19 in the real world.

In addition, at an effective infection number R_tIntroducing a dynamic parameter theta such that R_tCan exponentially decline with time, is matched with the exponential decline trend of the infection speed of the COVID-19 with time, and can better simulate the propagation condition of the COVID-19 in the real world. Wherein the effective infection number R_tIs shown in equation (13):

R_t＝R_∞+R₀*e^-θ*t(9)

wherein R is_tEffective number of infections, R, representing date t₀Denotes the number of primary infections initially (t ═ 0) without any external intervention, R_∞Represents COVIThe expected effective infection count at the end of D-19, and θ represents the rate of decline of the effective infection count. The parameter θ is affected by the following factors: when the prevention and control consciousness of people and the control policy of the government are strong, the decrease rate of the effective infection number is faster.

When the effective infection number R of COVID-19 is determined_tThe incubation period TE, the isolation period TI, the exposure state data (incubation period number E (0)) and the infection state data (infection period number I (0)) at the initial stage of epidemic outbreak are analyzed by an ordinary differential equation to obtain the results of the D-SEIQ model (S (t), E (t), I (t) and Q (t) at any moment), and the number of confirmed persons, the number of isolated persons and the expected inflection point at any moment can be calculated according to the results of the D-SEIQ model.

Wherein the parameters in the D-SEIQ model need to meet the range of medical rationality, as shown in table 1:

TABLE 1

Parameter(s)	Range of
		R₀	[2,7]
TE	[3,11]
		TI	[1,5]
R_∞	[0.05,0.4]
		θ	[0.05,0.45]

With the predefined parameter range, the method shown in fig. 6 can be used for parameter learning and epidemic situation trend prediction. As shown in fig. 6, the specific process of the D-SEIQ model is as follows: 1) collecting newly-increased diagnosis case number and accumulated diagnosis case number of COVID-19 reported by an official party; 2) setting an initial parameter R of a D-SEIQ model₀、TE、TI、R_∞Theta; 3) calculating the time-decaying R according to the exponential function of equation (9)_t(ii) a 4) Solving daily I (t), E (t) and Q (t) through ordinary differential equations (formulas (5) to (8)), and calculating the accumulated number of confirmed cases fitted by the D-SEIQ model through I (t), E (t) and Q (t); 5) calculating a mean square error MS E according to the accumulated number of confirmed cases and the real accumulated number of confirmed cases fitted by the D-SEIQ model, wherein,

f_icumulative number of confirmed cases, y, representing the fit_iThe number of actual cumulative confirmed cases is shown, and N shows the total days corresponding to the number of actual cumulative confirmed cases; 6) when the mean square error MSE is smaller than a set threshold value, the D-SEIQ model obtained at the date t is used as an optimal D-SEIQ model, and the long-term epidemic situation trend of COVID-19 is predicted based on the D-SEIQ model; 7) when the mean square error MSE is not minimum, namely the mean square error MSE is larger than or equal to a set threshold value, carrying out optimal parameter search on the real accumulated confirmed case number and the fitted accumulated confirmed case number through a heuristic search method based on a least square method so as to update the parameter R₀、TE、TI、R_∞And theta until an optimal D-SEIQ model is obtained.

The main data of the D-SEIQ model learning is the number of newly added confirmed cases every day, and as epidemic outbreak time, situation, detection capability and medical resource quantity of the areas B, A and A provinces are different, the three areas can be respectively modeled. Wherein the clinical diagnosis case can be adjusted prior to modeling: in the area of province A, because clinical diagnosis cases are included in the diagnosis standard from No. 02-12 to No. 02-14, newly added cases in the area of province A are greatly increased in three days, and newly added cases in the city B are increased to 13436 cases on the day of No. 02-12, wherein 12364 cases are all clinical diagnosis cases. The inclusion of clinical diagnosis cases has a great influence on the originally established D-SEIQ model. Therefore, dynamic adjustment of a sliding window is carried out on the historical newly-added data from No. 02-12 to No. 02-14, so that the historical newly-added cases are in accordance with the real situation, and the D-SEIQ model is retrained by using the adjusted data. The specific adjusting method comprises the following steps: clinical diagnosis cases from No. 02-12 to No. 02-14 are mainly suspected accumulated cases in the past 7-10 days, the area of province A except city B is 7 days, and the area of city B is 10 days, so the number of newly added suspected cases in the past 7-10 days in the area of province A is calculated, the percentage of the newly added suspected cases in each day is calculated, and clinical diagnosis cases from No. 02-12 to No. 02-14 are added to the newly added cases in the past 7-10 days to realize data adjustment.

In this case, the early data is adjusted so that the number of new cases per day in the D-SEIQ model increases gradually in the early stage of an infection outbreak, but in reality, the report number deviates from the number of patients due to the detection capability and the like. The areas of province A and city B are removed, the number of newly-increased daily cases from No. 01-20 to No. 01-23 is respectively 12, 0, 7 and 35, the number of newly-increased daily cases from No. 01-20 to No. 01-24 in the areas of city B is respectively 60, 105, 62, 70 and 77, and the newly-increased daily cases do not accord with the monotone increasing trend, so that exponential function fitting can be carried out on the early-stage case data of the two areas, and the D-SEIQ model is constructed after daily reported cases are redistributed.

In conclusion, the D-SEIQ model provided by the embodiment of the invention can simulate the process that a latent patient may infect other people in the COVID-19 infection process; the D-SEIQ model is combined with a parameter fitting technology, so that the model parameters can be better fitted by using a machine learning technology in the infection simulation process; the D-SEIQ model can accurately predict the long-term development trend of COVID-19 and the date range of epidemic inflection points in the region by using a small amount of COVID-19 data; the D-SEIQ model adopts a parameter fitting method to train partial parameters (R)₀、TE、TI、R_∞θ) capable of solving overfittingThe problem is solved, and meanwhile, the D-SEIQ model does not predict the infection trend based on machine learning, so that the problem of the machine learning model which is not interpretable can be solved.

To test the accuracy of the D-SEIQ model, the optimal D-SEIQ model was verified using the diagnosis data of the nationwide area other than the area a (the report data is less affected by human factors) to obtain the schematic diagram of the predicted cumulative confirmed cases nos. 01 to 26 shown in fig. 7A, the schematic diagram of the predicted daily newly-increased confirmed cases nos. 01 to 26 shown in fig. 7B, the schematic diagram of the predicted cumulative confirmed cases nos. 01 to 27 shown in fig. 8A, the schematic diagram of the predicted daily newly-increased confirmed cases nos. 01 to 27 shown in fig. 8B, the schematic diagram of the predicted cumulative confirmed cases nos. 02 to 04 shown in fig. 9A, the schematic diagram of the predicted daily newly-increased confirmed cases nos. 02 to 04 shown in fig. 9B, and the schematic diagram of the predicted cumulative confirmed cases nos. 02 to 11 shown in fig. 10A, Fig. 10B is a schematic diagram of newly diagnosed daily cases predicted to be newly diagnosed as nos. 02 to 11, fig. 11A is a schematic diagram of newly diagnosed daily cases predicted to be newly diagnosed as nos. 02 to 23, and fig. 11B is a schematic diagram of newly diagnosed daily cases predicted to be newly diagnosed as nos. 02 to 23. The schematic diagram of the cumulative confirmed cases is a trend diagram representing the trend of the cumulative confirmed cases, the abscissa of the schematic diagram of the cumulative confirmed cases is time, the ordinate is the number of the cumulative confirmed cases, and the number of the cumulative confirmed cases in the future can be clearly seen according to the schematic diagram of the cumulative confirmed cases; the schematic diagram of the newly-added diagnosed cases every day is a trend diagram representing the trend of the newly-added diagnosed cases, the abscissa of the schematic diagram of the newly-added diagnosed cases every day is time, and the ordinate is the number of the newly-added diagnosed cases every day. The prediction result of the D-SEIQ model No. 01-27 shows that the number of accumulated confirmed cases of epidemic situation reaches about 12304 in No. 02-23, and the peak of newly-increased confirmed cases in No. 02-01 reaches about 780 every day. Compared with real data, the estimated data of the D-SEIQ model for 1 month is very close to the real data, wherein the number 02-23 of real accumulated numbers is 12863, and the difference between the number 02 and the number 23 and the predicted value is only 5%. The inflection point of the epidemic situation predicted by the D-SEIQ model is No. 02-01 and is consistent with the real date (between No. 02-01 and No. 02-03). Therefore, in the areas of provinces A all over the country, the D-SEIQ model can estimate the trend of up to 1 month very accurately in the early stage (about one week) of epidemic outbreak. D-SEIQ model results of late epidemic situation transmission show that D-SEIQ models of No. 02-11 and No. 02-23 both have relatively accurate fitting on real data, the number of accumulated confirmed cases of No. 23 in No. 2 months is estimated to be 12832 by the D-SEIQ model of No. 02-11, the number of real confirmed cases is 12863, and only 31 persons are different. The embodiment of the invention is not limited to areas except provinces A in the whole country.

In addition, the embodiment of the present invention is directed to the effective infection number R_tFitting analysis is carried out, and the change trend of the effective infection number of three regions (B city, A province except B city, and nationwide except A province) is obtained. Using the viral spread parameters learned by the late epidemic model as estimates of the true model parameters, curve fitting was performed on the basic spread numbers from 01-20 to 03-10, as shown in FIG. 12, where A province and B provinces are divided nationally by R ₀6, all of which are all except for R in the regions of province A and the regions of city B₀R in the region greater than B₀. However, the R in the regions of provinces A is removed from the whole country₀The descending rate (0.20) is greater than R of province A, city B₀Rate of decrease (0.15), A omits R in B₀The descending rate is greater than R of the region in the city B₀The rate of decrease (0.10). R of regions of province A across the country_tFirst, the number of the main points is reduced to less than 1, and then, the province A excludes the region of city B and the region of city B respectively. Ultimate R in three regions_∞The values all tend to zero, which indicates that the prevention and control measures are effective in time.

Therefore, by summarizing the parameters of the virus properties learned by the D-SEIQ model, the following conclusions are drawn: 1) from the early stage of epidemic situation, if strong and effective control measures (closing a city, wearing a mask, isolating, shutting down, screening) and the like are not carried out, the epidemic situation will outbreak nationwide in a short period because the basic infection ratio of the new coronary pneumonia epidemic virus is relatively large and is about 3-5; 2) from retrospective data, the development of epidemic situations is effectively restrained by control measures all over the country, the effective infection number of the final epidemic situations is reduced to about 0.2, namely 1 normal person can be effectively infected by 5 patients in the late stage of the epidemic situations, and the control effects on the spread of the epidemic situations are remarkable by public control awareness (wearing a mask, in a home and in a gathering) and government epidemic control measures (closing a city, shutting down and isolating); 3) the prevention and control measures (No. 01-27) in areas except provinces A in China show an effect in about 1 week of epidemic outbreak, so that the D-SEIQ model can accurately predict the epidemic trend of nearly 1 month in early stage of the epidemic, the difference between the accumulated numbers of No. 02-23 is only 5%, and the turning point of the epidemic is consistent with the real situation (about No. 02-01); 4) because the region in the city B is in an epidemic outbreak center, the number of cases of the epidemic, the shortage of early detection reagents and the relative shortage of medical resources are caused, so that the basic propagation number of the region in the city B is far larger than that of other regions, the isolation period time TI is also larger than that of other regions, and the inflection point appears late (about No. 02-09).

Now, the method for building an infectious disease trend prediction model according to the embodiment of the present invention has been described with reference to the exemplary application and implementation of the server according to the embodiment of the present invention, and the following continues to describe a scheme for implementing building an infectious disease trend prediction model by matching the modules in the apparatus 555 for building an infectious disease trend prediction model according to the embodiment of the present invention.

An attenuation module 5551 for attenuating the basic infection count of the infectious disease according to the transmission time to obtain effective infection counts for a plurality of dates during the transmission period; a determining module 5552, configured to determine fitted state data of a plurality of dates corresponding to the effective infection numbers of the plurality of dates in a state transition relationship of the infectious disease included in the infectious disease trend prediction model; an extracting module 5553, configured to extract fitting case data of the multiple dates from the fitting state data of the multiple dates; an updating module 5554, configured to update parameters of the infectious disease trend prediction model according to a difference between the real case data of multiple dates and the fitted case data of multiple dates, and use the updated parameters as parameters used for predicting an infectious disease epidemic situation trend based on the infectious disease trend prediction model.

In some embodiments, the attenuation module 5551 is further configured to add the expected effective infection counts at the end of the infection and the attenuation results corresponding to the basic infection counts on the dates to obtain effective infection counts on the dates; wherein the attenuation result is obtained by attenuating the basic infection number based on the transmission time of the infectious disease.

In some embodiments, the decay module 5551 is further configured to determine an effective infection count for any of the plurality of dates t according to the following formula: r_t＝R_∞+R₀*e^-θ*tWherein R is_tRepresents the effective infection number, R, of said date t₀*e^-θ*tRepresenting the result of the decay of said primary infection count at said date t, R₀Representing said base infection number, theta representing the rate of decline of said effective infection number, R_∞Indicating the expected effective number of infections at the end of the infection.

In some embodiments, the determining module 5552 is further configured to determine, as fitting status data of the plurality of dates, status data that corresponds to the plurality of dates one by one and satisfies the constraint condition, with the status conversion relationship of the infectious diseases as a constraint condition and the effective infectious counts of the plurality of dates as known quantities in the constraint condition; wherein, the state transition relation of the infectious disease comprises parameters of the infectious disease trend prediction model.

In some embodiments, the state of the infectious disease comprises a susceptible state, an exposed state, an infected state, and an isolated state; the infectious disease trend prediction model comprises the state transition relationships including: a conversion relationship between the susceptible state of date t, the infectious state of date t, an infection rate of the infectious disease, and a rate of decline of the susceptible state of date t, a conversion relationship between the susceptible state of the date t, the infectious state of the date t, the exposed state of the date t, the infection rate of the infectious disease, the incidence rate of the infectious disease, and the rising rate of the exposed state of the date t, a conversion relationship between the exposure state of the date t, the infection state of the date t, the incidence rate of the infectious disease, the quarantine rate of the infectious disease, and the rising rate of the infection state of the date t, a conversion relationship between the infection state of the date t, an isolation speed of the infectious disease, and a rising rate of the isolation state of the date t; the fitting state data comprises fitting susceptible state data, fitting exposed state data, fitting susceptible pathological state data and fitting isolated state data.

In some embodiments, the state transition relationship of infectious disease comprises:

wherein, s (t) + e (t) + i (t) + q (t) ═ N, infection rate β ═ R_t(t) fitting susceptibility data on date t, (e) fitting exposure data on date t, (i) (t) fitting susceptibility data on date t, (q) (t) fitting isolation data on date t, and (R) isolation rate γ 1/TI_tEffective infection number representing date t, TE representing a latency period of the infectious disease, TI representing an isolation period of the infectious disease, and the fitting state data comprising the fitting susceptibility state data, the fitting exposure state data, the fitting susceptibility state data, and the fitting isolation state data.

In some embodiments, the fitted case data comprises a fitted diagnosed case; the extracting module 5553 is further configured to extract fitting exposure state data, fitting susceptibility state data, and fitting isolated state data corresponding to the multiple dates one by one from the fitting state data of the multiple dates, add the fitting exposure state data, the fitting susceptibility state data, and the fitting isolated state data corresponding to the multiple dates one by one, and use a sum result as a fitting diagnosed case corresponding to the multiple dates one by one.

In some embodiments, the update module 5554 is further configured to construct an error function of the infectious disease trend prediction model based on differences between real case data for multiple dates of the infectious disease and fitted case data for the multiple dates; and updating the parameters of the infectious disease trend prediction model until the error function converges.

In some embodiments, the parameters of the infectious disease trend prediction model include a baseline infection count, a latency period, an isolation period, an expected effective infection count at the end of an infectious disease, and a rate of decline of the effective infection count of the infectious disease; the updating module 5554 is further configured to determine an updating direction of the basic infection number, the latency period duration, the isolation period duration, the effective infection number at the end of the infectious disease, and the decreasing rate of the effective infection number in the infectious disease trend prediction model when the value of the error function is greater than an error function threshold, and update the basic infection number, the latency period duration, the isolation period duration, the effective infection number at the end of the infectious disease, and the decreasing rate of the effective infection number according to the updating direction until the error function satisfies a convergence condition.

Now, the method for predicting the tendency of the infectious disease based on the infectious disease tendency prediction model according to the embodiment of the present invention is described, and the scheme for cooperatively realizing the prediction of the tendency of the infectious disease based on the infectious disease tendency prediction model by each module in the device 556 according to the embodiment of the present invention is described.

A first determination module 5561, configured to attenuate the basic infection count of the infectious disease according to the transmission time, and obtain effective infection counts corresponding to a plurality of predicted dates of the infectious disease; a second determination module 5562, configured to determine, as fitting state data of the plurality of predicted dates, state data that respectively correspond to the plurality of predicted dates and satisfy the constraint condition, using the state transition relationship included in the infectious disease trend prediction model as a constraint condition, and using the effective infection counts of the plurality of predicted dates as known quantities in the constraint condition; the processing module 5563 is configured to sum the fitted exposed state data, the fitted infection state data, and the fitted isolated state data in the fitted state data of the multiple prediction dates to obtain fitted confirmed cases of the multiple prediction dates, and generate an epidemic situation trend formed by the infectious disease in the multiple prediction dates according to the number change of the fitted confirmed cases of the multiple prediction dates.

In some embodiments, the first determining module 5561 is further configured to add the expected effective infection counts at the end of the infection and the attenuation results of the basic infection counts at the plurality of predicted dates to obtain effective infection counts at the plurality of predicted dates;

In some embodiments, the state of the infectious disease comprises a susceptible state, an exposed state, an infected state, and an isolated state; the second determining module 5562 is further configured to substitute the effective infection counts for the plurality of predicted dates in the following formula of the state transition relationship to obtain the fitted exposure state data, the fitted infection state data, and the fitted isolation state data for the plurality of predicted dates:

Embodiments of the present invention also provide a computer-readable storage medium storing executable instructions, which when executed by a processor, will cause the processor to execute the method for constructing an infectious disease trend prediction model provided by embodiments of the present invention, for example, the method for constructing an infectious disease trend prediction model shown in fig. 3A-3B, and the method for predicting an infectious disease epidemic situation trend based on an infectious disease trend prediction model shown in fig. 4.

In some embodiments, the computer-readable storage medium may be memory such as FRAM, ROM, PROM, EP ROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or may be various devices including one or any combination of the above memories.

In some embodiments, executable instructions may be written in any form of programming language (including compiled or interpreted languages), in the form of programs, software modules, scripts or code, and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

By way of example, executable instructions may correspond, but do not necessarily have to correspond, to files in a file system, may be stored in a portion of a file that holds other programs or data, e.g., in one or more scripts in a HyperText markup Language (H TML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub-programs, or portions of code).

By way of example, executable instructions may be deployed to be executed on one computing device (a device that includes a smart terminal and a server), or on multiple computing devices located at one site, or distributed across multiple sites and interconnected by a communication network.

In summary, the embodiment of the invention predicts the epidemic situation trend based on the infectious disease trend prediction model, and has the following beneficial effects:

1. dynamic parameters are introduced into the effective infection number, so that the effective infection number can exponentially decline along with time, and the trend that the effective infection number in real infectious diseases exponentially declines along with time is fitted;

2. the infectious disease trend prediction model is combined with a parameter fitting technology, so that the model parameters can be better fitted by using a machine learning technology in the infection simulation process, the overfitting problem is solved, and meanwhile, the D-SEIQ model does not predict the infectious disease trend based on machine learning, so that the problem that the machine learning model cannot be explained can be solved;

3. the infectious disease trend prediction model can accurately predict the long-term development trend of the infectious disease and the date range of epidemic inflection points in the region by using a small amount of infectious disease data;

4. the infectious disease trend prediction model adopts a parameter fitting method to train partial parameters (R)₀、TE、T I、R_∞θ), and the infection tendency prediction model does not predict the infection tendency based on machine learning, so that the problem of inexplicability of the machine learning model can be solved.

The above description is only an example of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, and improvement made within the spirit and scope of the present invention are included in the protection scope of the present invention.

Claims

1. A method for constructing an infectious disease trend prediction model is characterized in that,

the infectious disease trend prediction model is used for predicting epidemic situation trends of infectious diseases;

the method comprises the following steps:

2. The method of claim 1, wherein attenuating the baseline transmission of the infectious disease by transmission time to obtain an effective transmission number for a plurality of dates during the transmission period comprises:

adding the expected effective infection counts at the end of the infection and the attenuation results corresponding to the basic infection counts on the plurality of dates to obtain effective infection counts on the plurality of dates;

3. The method according to claim 2, wherein the adding the expected effective infection counts at the end of the infectious disease and the attenuation results of the basic infection counts at the plurality of dates to obtain the effective infection counts at the plurality of dates, respectively, comprises:

determining the effective infection count of any one of the plurality of dates t according to the following formula:

4. The method of claim 1, wherein the determining fitted state data of a plurality of dates in one-to-one correspondence with the effective infection counts of the plurality of dates in the state transition relationships of the infectious disease included in the infectious disease trend prediction model comprises:

determining state data which corresponds to the dates one by one and meets the constraint condition by taking the state conversion relation of the infectious diseases as a constraint condition and taking the effective infectious counts of the dates as known quantities in the constraint condition, wherein the state data are used as fitting state data of the dates;

5. The method of claim 4,

the states of the infectious disease include a susceptible state, an exposed state, an infected state and an isolated state;

6. The method according to claim 4 or 5, wherein the state transfer relationship of infectious diseases comprises:

7. The method of claim 1,

the fitted case data comprises fitted confirmed cases;

the extracting of the fitting case data of the plurality of dates from the fitting state data of the plurality of dates includes:

extracting fitting exposed state data, fitting susceptible state data and fitting isolated state data which are in one-to-one correspondence with the dates from the fitting state data of the dates, and

8. The method of claim 1, wherein updating parameters of the infectious disease trend prediction model based on differences between real case data for multiple dates of the infectious disease and fitted case data for the multiple dates comprises:

constructing an error function of the infectious disease trend prediction model according to the difference between the real case data of the infectious disease on a plurality of dates and the fitted case data of the plurality of dates;

9. The method of claim 8,

the parameters of the infectious disease trend prediction model comprise basic infection number, latent period duration, isolation period duration, expected effective infection number at the end of infectious disease and the descending rate of the effective infection number of the infectious disease;

the updating the parameters of the infectious disease trend prediction model until the error function converges comprises:

when the value of the error function is larger than the threshold value of the error function, determining the updating directions of the basic infection number, the latent period duration, the isolation period duration, the effective infection number at the end of the infectious disease and the descending rate of the effective infection number in the infectious disease trend prediction model, and determining the updating directions

10. An infectious disease epidemic situation trend prediction method based on an infectious disease trend prediction model is characterized by comprising the following steps:

11. The method of claim 10, wherein attenuating the baseline infection count for the infectious disease based on time of transmission to obtain an effective infection count for a plurality of predicted dates of the infectious disease comprises:

adding the expected effective infection counts at the end of the infectious disease and the attenuation results of the basic infection counts corresponding to the plurality of prediction dates to obtain effective infection counts of the plurality of prediction dates;

12. The method of claim 10,

the method for determining the state data which respectively correspond to the plurality of prediction dates and satisfy the constraint condition by using the state conversion relation included in the infectious disease trend prediction model as the constraint condition and using the effective infection numbers of the plurality of prediction dates as the known quantity in the constraint condition as the fitting state data of the plurality of prediction dates comprises the following steps:

substituting the effective infection numbers of the plurality of prediction dates into the formula of the state conversion relation to obtain the fitted exposure state data, the fitted infection state data and the fitted isolation state data of the plurality of prediction dates:

13. An apparatus for constructing an infectious disease trend prediction model, the apparatus comprising:

the device comprises:

14. An electronic device, characterized in that the electronic device comprises:

a memory for storing executable instructions;

a processor for implementing the method of constructing an infectious disease trend prediction model of any one of claims 1 to 9 when executing executable instructions stored in the memory.

15. A computer-readable storage medium storing executable instructions for causing a processor to implement the method for constructing an infectious disease trend prediction model according to any one of claims 1 to 9 or the method for predicting an infectious disease epidemic situation based on an infectious disease trend prediction model according to any one of claims 10 to 12 when the processor executes the executable instructions.