WO2021189516A1

WO2021189516A1 - Method and system for simulating process of temporal and spatial circulation of influenza with massive trajectory data

Info

Publication number: WO2021189516A1
Application number: PCT/CN2020/082708
Authority: WO
Inventors: 尹凌; 张�浩
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2020-03-27
Filing date: 2020-04-01
Publication date: 2021-09-30
Also published as: CN113450923B; CN113450923A

Abstract

A method and a system for simulating the process of temporal and spatial circulation of influenza with massive trajectory data. Said method comprises: synthesizing an urban population on the basis of population census data and building census data, and assigning a corresponding population attribute to an individual of the synthesized urban population (S1); for the individual of the synthesized urban population to which the population attribute has been assigned, constructing an individual activity chain mainly consisting of mobile phone position data and being assisted by travel investigation data (S2); on the basis of the constructed individual activity chain, dynamically constructing, with one hour as a time step, a contact network of 24 time series in a day (S3); according to the constructed contact network, using an SEIR model to simulate the circulation of an influenza at a high temporal and spatial resolution (S4); and analyzing a simulation result from two perspectives of time and space, so as to analyze a high-risk area during circulation of the influenza and accurately position a key spatial position during circulation of influenza epidemic situation (S5). Said method can invert the temporal and spatial information of influenza outbreaks on different temporal and spatial scales, effectively solving the problem of being unable to realize real location of the spatial position of an infected person and a potentially high-risk circulation area.

Description

Method and system for simulating the spreading process of influenza in time and space with large-scale trajectory data

Technical field

The invention relates to a method and a system for simulating the time-space propagation process of influenza with large-scale trajectory data.

Background technique

Influenza is an acute respiratory infection caused by influenza virus and parainfluenza virus. Through the droplets and particles produced when the infected person coughs or sneezes, the virus can easily spread from person to person and seriously endanger the lives and health of the people. . The temporal and spatial evolution of human activities and related fluctuations in population density are the key driving factors for the dynamics of infectious disease outbreaks. By tracing the contact between urban individuals at high spatial and temporal resolution, starting from discrete individuals, inferring potentially infected individuals based on the time and location information of the activities of different individuals, it helps to accurately express the spatiotemporal spread of influenza and improve the spread of influenza Accuracy of spatial model prediction. In recent years, there has been an explosive growth in large-scale individual trajectory data (mobile phone location data, floating car GPS data, bus card swiping data, etc.), which can accurately locate the time and space information of mobile individuals, in order to break through the time and space accuracy of infectious disease prevention and control. Bottlenecks provide new opportunities.

At present, the agent-based infectious disease model at the urban scale mainly reconstructs the movement of individuals and the contact network between individuals through travel survey data, and studies the transmission characteristics of infectious diseases in time series.

However, the prior art has at least the following shortcomings: building a real-world oriented agent model requires a large amount of real individual data, otherwise it is impossible to know precise individual positions and contacts between individuals. Although existing studies have attempted to model population movement based on trajectory data to build an infectious disease spread model, the existing trajectory data-based infectious disease spread modeling method still cannot effectively solve the integration of population attributes (such as age, gender, occupation, etc.). Family structure) and individual modeling needs of the movement characteristics of trajectory data, lack of effective methods for fusing trajectory data to construct temporal and spatial proximity relationships of urban individuals (for example, individuals appearing in the same place at the same time), and even neglecting individuals in densely populated modern cities The complexity of the contact space; at the same time, the trajectory data has sample bias and cannot represent the entire population of the entire city. In view of the spread of infectious diseases, individuals show obvious heterogeneity in the spread of influenza viruses. Many existing studies ignore the mutual influence of parameters in order to simplify the model parameters. In addition, agent-based spatial explicit infectious disease models are often studied at low temporal and spatial resolutions, and it is difficult to reveal the spread of influenza at the urban scale from the perspective of temporal and spatial dynamic mechanisms, resulting in delays and delays in predicting the temporal and spatial patterns of influenza outbreaks. deviation.

Summary of the invention

In view of this, it is necessary to provide a method and system for simulating the spread of influenza in time and space with large-scale trajectory data.

The present invention provides a method for simulating the spatiotemporal transmission process of influenza with large-scale trajectory data. The method includes the following steps: a. Synthesize urban population based on census data and building census data, and assign corresponding population attributes to individuals of the synthesized urban population B. For the individuals of the synthetic urban population given demographic attributes, use mobile phone location data as the main and travel survey data as a supplement to construct an individual activity chain; c. Based on the constructed individual activity chain, one hour is the time Step size, dynamically construct a contact network of 24 time series in a day; d. According to the constructed contact network of 24 time series in a day, use the SEIR model to simulate the spread of influenza at high spatial and temporal resolution.

Wherein, the method also includes the steps: the method also includes the step e: analyzing the simulation results from the two perspectives of time and space to obtain the spatial transmission path between the infected persons, and accurately locate the key spatial locations in the spread of the influenza epidemic .

The population census data includes: age, gender, occupation type, family category, family size, and family age components; the building census data includes: building location information, building height, building area, and building function; wherein, the Building functions include: factories, teaching buildings, residential buildings, office buildings, and shopping malls.

The demographic attributes include individual attributes, family attributes, and whether they are individuals with mobile phones; wherein, the individual attributes include age, gender, and occupation; the family attributes include family structure, home address, and work place.

Said step b specifically includes the following steps: constructing the travel trajectory of an individual with a mobile phone based on mobile phone location data: sorting the data of the same mobile phone number by time to form a day’s travel trajectory of the mobile phone user, and dividing Tyson based on the mobile phone base station Polygon, multiple buildings located in the same Tyson polygon as the mobile phone base station are the candidate sets of individual positions; the travel trajectory of individuals without mobile phones is constructed based on travel survey data: the same The multiple buildings in the traffic area are candidate sets of individual locations; according to the candidate set of individual locations obtained above, an individual activity chain is constructed.

Said step c specifically includes the following steps: according to the activity chain of the constructed individual, with the hour as the time granularity, compare the activity categories performed by different individuals in a day, and set individuals who perform the same activities at the same time and are in the same building location For individuals with co-occurrence in time and space, based on the activity category, different contact probabilities are assigned to individuals in co-occurrence in time and space. Under the constraint of contact probability, a dynamic contact network with hourly resolution between agents is generated.

The activity categories include: home, work/school, leisure and entertainment activities.

The step d specifically includes the following steps: using hours as the unit of time, tracking the time and place of the infection, and from whom to whom. During the spread of influenza, the probability of an individual being infected is called the effective infection probability P, The formula for effective infection probability is as follows:

P=P _c ×P _i ×r

Among them, P _c is the contact probability between individuals, P _i is the infection probability of the individual, and r is the relative infectivity of the individual;

Finally, the Monte Carlo method is used to determine whether the individual is infected.

The Monte Carlo method to determine whether the individual is infected includes: generating a uniformly distributed pseudo-random number based on a computer, comparing the pseudo-random number with the effective infection probability, and if the pseudo-random number is less than or equal to the effective infection probability, the individual is infected , Repeat the above process until the final trend of the number of new infections per day is consistent with the real data and the calculated basic reproductive number is greater than 1.

The step e specifically includes the following steps: a comparative analysis of the city scale and the administrative division scale on the time series, divide the city into 1km×1km grids, and analyze the number of newly infected cases in each grid every day ；

The epidemic tree is constructed by constructing the spread topological relationship between the parent infected person and the offspring infected person, and then the epidemic forest is composed of multiple epidemic trees to obtain the spatial transmission path between the parent infected person and the offspring infected person, and accurately locate the flu The key spatial location in the spread of the epidemic.

The present invention provides a system for simulating the spatiotemporal transmission process of influenza with large-scale trajectory data. The system includes a population attribute assignment module, an activity chain building module, a contact network building module, and a transmission simulation module. The census data and building census data synthesize the urban population, and assign corresponding population attributes to the individuals of the composite urban population; the activity chain building module is used to assign the population attributes to the individuals of the composite urban population, mainly based on mobile phone location data , The travel survey data is supplemented to construct an individual's activity chain; the contact network building module is used to dynamically construct a contact network of 24 time series in a day based on the constructed individual's activity chain, taking one hour as the time step The spread simulation module is used to simulate the spread of influenza at high temporal and spatial resolution by using the SEIR model according to the constructed contact network of 24 time series in a day.

The beneficial effects of the method and system for simulating the time-space propagation process of influenza with large-scale trajectory data of the present invention include:

(1) With the integration of large-scale mobile phone location data, the mobility of individuals is more realistically restored, the mobility of regions is more concentrated, and the activity space has increased compared with travel survey data. The travel position of the individual in the mobile model is represented by building coordinates, so that the spatial explicit infectious disease model based on the agent can retrieve the spatiotemporal information of influenza outbreaks on different spatiotemporal scales.

(2) The agent-based spatial explicit epidemic model sets a large number of parameters related to the spread of influenza, improves the interpretability of the model, and helps to reveal the spreading mechanism of influenza at high spatial and temporal resolution. The simulation results can match the trend of the original influenza data at the urban scale and administrative division scale. The spatial information of the epidemic situation described by the simulation results effectively solves the problem of the inability to truly locate the spatial location of the infected person and the potential high-risk transmission area.

(3) Innovatively use epidemic forests to characterize the spread of influenza in a small-scale space, analyze the transmission characteristics of influenza within and between grids, and locate high-risk areas in the process of influenza transmission. The model can reflect the intensity of the spread of influenza within the region, depict the law of spread of influenza between regions, and help to accurately locate the key spatial locations in the spread of influenza epidemics.

Description of the drawings

Fig. 1 is a flow chart of the method for simulating the temporal and spatial propagation process of influenza with large-scale trajectory data according to the present invention;

Figure 2 is a schematic diagram of the natural history of influenza under the SEIR model;

Figure 3 is a schematic diagram of an epidemic forest provided by an embodiment of the present invention;

Figure 4 is a hardware architecture diagram of the system for simulating the spread of influenza in time and space with large-scale trajectory data according to the present invention;

5 is a schematic diagram of the age and gender distribution of synthetic individuals provided by an embodiment of the present invention;

FIG. 6 is a schematic diagram of the structure distribution of a synthetic family provided by an embodiment of the present invention;

FIG. 7 is a schematic diagram of the comparison between the simulation results of the influenza diffusion process at the urban scale and the real cases according to the embodiment of the present invention;

FIG. 8 is a schematic diagram of the distribution of reproduction numbers generated by 100 simulation results at a city scale according to an embodiment of the present invention;

Fig. 9 is a schematic diagram of the comparison between the simulation results of the influenza spreading process provided by the embodiment of the present invention and real cases on a scale of 10 regions;

FIG. 10 is a schematic diagram of the temporal and spatial distribution of influenza spread in a city according to an embodiment of the present invention;

FIG. 11 is a schematic diagram of the intensity of the spread of influenza within the space unit according to an embodiment of the present invention;

FIG. 12 is a schematic diagram of the number of grids that can be affected by one grid during the spread of influenza according to an embodiment of the present invention;

FIG. 13 is a schematic diagram of the number of events (connection strength) in which influenza transmission occurs between two grids according to an embodiment of the present invention;

FIG. 14 is a schematic diagram of a spatial aggregation phenomenon during the spread of influenza according to an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions, and advantages of this application clearer and clearer, the following further describes the application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the application, and are not used to limit the application.

Referring to FIG. 1, it is a flowchart of a preferred embodiment of the method for simulating the time-space propagation process of influenza with large-scale trajectory data according to the present invention.

Step S1: Synthesize the urban population based on the population census data and the building census data, and assign corresponding population attributes to individuals of the synthesized urban population. That is, fusion of multi-source data to build a model of urban population movement. in:

in particular:

First, according to the probability distribution of individual attributes such as age, gender, and occupation type in the census data, Monte Carlo simulation is used to assign corresponding individual attributes to the individuals of each composite urban population. Secondly, according to the probability of family attributes such as family category, family size, family age composition and so on in the census data, a synthetic family is constructed, and individuals of the synthetic urban population are filled into the synthetic family. Finally, according to the mobile phone usage rate of different genders and age groups, Monte Carlo simulation is performed to give the individuals of the synthetic urban population the attributes of whether they are individuals with mobile phones. The individuals of the synthetic urban population are divided into two types: individuals with mobile phones and individuals without mobile phones. .

Step S2, for individuals (hereinafter referred to as "individuals") of synthetic urban population assigned population attributes, using mobile phone location data as the main and travel survey data as the supplement to construct an individual activity chain. That is to say, based on the combination of mobile phone data and travel survey data supplemented by the time and space characteristics of individual travel and building census data, building-scale individual movement modeling is realized. in particular:

The travel trajectory of individuals with mobile phones is reconstructed based on mobile phone location data, and the travel trajectories of individuals without mobile phones are constructed based on travel survey data. At this time, the spatial range of the work and residence of individuals with mobile phones identified through the mobile phone data is refined to the service range of the mobile phone base station, and the spatial range of the work and residence of individuals without mobile phones based on travel survey records is refined to the traffic community.

Further, the mobile phone location data includes: anonymous mobile phone number, time, base station latitude and longitude. The data of the same mobile phone number is sorted by time to form a day's travel trajectory of the mobile phone user. At this time, the user's travel location is the location of the mobile phone base station. . Then, based on the mobile phone base station, the Thiessen polygon is divided (the range of the Thiessen polygon is the service range of the mobile phone base station), and multiple buildings located in the same Thiessen polygon as the mobile phone base station are the candidate sets of individual positions.

Travel survey is an investigation of individual travel behavior. The travel survey data includes: individual work unit, place of departure, destination, departure time, end time, travel mode, travel purpose, etc., but the location information is based on the traffic community. The travel trajectory of an individual without a mobile phone is constructed with each travel information in the travel survey, and multiple buildings in the same traffic area are obtained as a candidate set of individual locations.

According to the candidate set of individual positions obtained above, an individual's activity chain is constructed.

Step S3, based on the constructed individual activity chain, with one hour as the time step, dynamically constructing a contact network of 24 time series in a day.

That is, a contact network is dynamically constructed based on the individual activity chain of the urban population. According to the constructed activity chain of individuals, with the hour as the time granularity, compare the types of activities performed by different individuals within a day, and set individuals who perform the same activities at the same time and are in the same building location as individuals with the same time and space. Based on the activity category, different contact probabilities are assigned to individuals who co-occur in time and space. Under the constraint of contact probability, a dynamic contact network with hourly resolution between agents is generated. Wherein, the activity categories include: home, work/school, leisure and entertainment activities.

in particular:

In a contact network in a period of time, the vertices represent traveling individuals, and individuals appearing in the same location are connected by edges. The locations are in units of home (home), work (work), and building address (leisure). Individuals appearing at the same location at the same time have a certain probability of contact, and this probability of contact is recorded as p _c .

An individual has 24 contact networks in a day. Take a contact network as an example. The vertices in the network represent individuals. Individuals appearing at the same location are connected by edges, which means that two individuals travel at the same location at the same time. Then, according to the contact probability in Table 1, it is judged whether two individuals appearing at the same place at the same time actually have contact. For example, when the infected person is an adult, he will have contact with a minor at a probability of 0.25 when he is at home. 0.4 probability of contact with an adult.

When individuals are engaged in different activities, the contact probabilities between individuals are different (see Table 1). According to the contact probability, a contact network between individuals is generated every hour, and 24 contact networks a day constitute the final dynamic contact network.

Table 1 Probability of contact between individuals

Step S4: According to the constructed contact network of 24 time series in a day, the SEIR model is used to simulate the spread of influenza at high temporal and spatial resolution. in particular:

The SEIR model divides individuals into four states based on the natural history of influenza (Figure 2): susceptible period, incubation period, infection period, and recovery period. As shown in Figure 2, due to the vaccination and the production of autoantibodies, a part of the susceptible population has immunity to influenza virus; a susceptible individual is infected with a certain probability and enters the incubation period; after the virus is parasitic in the body for a certain period of time, the individual has Infectious, entering the infectious period; an infected person in the infectious period may show flu-corresponding symptoms, or there may be no obvious symptoms. Finally, the individual was cured and entered a recovery state.

The simulation process takes hours as the time unit to track when and where the infection occurred, and from whom to whom. During the spread of influenza, the probability of an individual being infected is called the effective probability of infection P. In the model, the formula for effective infection probability is as follows:

P=P _c ×P _i ×r

Wherein P _c is equal probability of contact between individuals, P _i is the probability of infected individuals, r is the relative infectivity for the individual. Finally, the Monte Carlo method is used to determine whether the individual is infected: based on a computer-generated uniformly distributed pseudo-random number, the pseudo-random number is compared with the effective infection probability, if the pseudo-random number is less than or equal to the effective infection probability, the individual is infected. Repeat the above process until the final trend of the number of new infections per day is consistent with the real data and the calculated basic reproductive number is greater than 1.

Step S5: Analyze the simulation results from the two perspectives of time and space, explore the trend and intensity of influenza outbreaks, analyze high-risk areas in the process of influenza transmission, and accurately locate key spatial locations in the process of influenza epidemic transmission.

Wherein, the influenza outbreak trend and intensity refers to: the influenza outbreak trend is mainly reflected in the time curve of newly infected cases every day, and the curve is roughly normal distribution. If the curve amplitude is narrow, it indicates that the outbreak speed is relatively high. Fast, on the contrary, the speed of the outbreak is relatively slow, the peak of the curve indicates the severity of the outbreak, the higher the peak, the more serious.

The high-risk areas are areas where the cumulative number of infected persons is large, which means that susceptible persons have a high risk of being infected in these areas.

The key spatial location is a location with strong transmission connectivity between regions, that is, intervention in the key spatial location can reduce the spread of influenza virus to other regions, thereby reducing the risk of transmission of infectious diseases within the city.

Specifically:

On the time series, a comparative analysis of the city scale and the administrative division scale is carried out. In order to accurately express the spread of influenza in time and space, the city is divided into 1km×1km grids, and the number of newly infected cases per grid per day is analyzed. The number of cases is the average of 100 simulation results. The epidemic tree is constructed by constructing the transmission topology relationship between the parent infected and the offspring infected, and then the epidemic forest is composed of multiple epidemic trees (Figure 3) to reveal the spatial transmission between the parent infected and the offspring infected Path, accurately locate the key spatial locations during the spread of influenza epidemics, reveal the relationship between the spread distance and the transmission intensity of influenza in the city, and the spatial agglomeration effect.

The key spatial position in the process of locating the influenza epidemic is mainly used to analyze the transmission intensity of influenza virus between spatial grids. The stronger the transmission connectivity between one grid and other grids (Figure 13), and the grid The greater the number of related grids (Figure 12), the more important the location is, and key intervention is required.

The spatial distance between the grids represents the transmission distance of influenza, and the connection weight between the grids represents the transmission intensity of influenza between the grids. In order to reveal the relationship between the transmission distance and the transmission intensity, the Pearson between the two is calculated in this embodiment. The correlation coefficient r, r=-0.098, shows that the propagation distance and the propagation strength are extremely weakly negatively correlated.

The spatial agglomeration effect is reflected in the obvious spatial agglomeration in areas where the connection weight is greater than 100 (Figure 14), indicating that these areas have a strong ability to communicate with each other.

Refer to FIG. 4, which is a hardware architecture diagram of the system 10 for simulating the time-space propagation process of influenza with large-scale trajectory data according to the present invention. The system includes: a population attribute assignment module 101, an activity chain building module 102, a contact network building module 103, a propagation simulation module 104, and an analysis module 105.

The population attribute assignment module 101 is used to synthesize urban population based on census data and building census data, and assign corresponding population attributes to individuals who synthesize urban population. That is, fusion of multi-source data to build a model of urban population movement. in:

in particular:

The population attribute assignment module 101 first assigns corresponding individual attributes to individuals of each composite urban population through Monte Carlo simulation based on the probability distribution of individual attributes such as age, gender, and occupation type in the census data. Then a composite family is constructed according to the probability of family attributes such as family category, family size, and family age composition in the census data, and the individuals of the composite urban population are filled into the composite family. Finally, Monte Carlo simulation is carried out according to the mobile phone usage rate of different genders and age groups, and the attributes of the synthetic urban population are assigned to individuals with mobile phones. The synthetic urban population individuals are divided into two categories: mobile phone individuals and mobile phone individuals.

The activity chain construction module 102 is used to construct an individual activity chain for individuals (hereinafter referred to as "individuals") of the synthetic urban population assigned population attributes, using mobile phone location data as the main and travel survey data as the supplement. That is to say, based on the combination of mobile phone data and travel survey data supplemented by the time and space characteristics of individual travel and building census data, building-scale individual movement modeling is realized. in particular:

The activity chain construction module 102 reconstructs the travel trajectory of individuals with mobile phones based on mobile phone location data, and constructs the travel trajectories of individuals without mobile phones based on travel survey data. At this time, the spatial range of the work and residence of individuals with mobile phones identified through the mobile phone data is refined to the service range of the mobile phone base station, and the spatial range of the work and residence of individuals without mobile phones based on travel survey records is refined to the traffic community.

Further, the mobile phone location data includes: anonymous mobile phone number, time, base station latitude and longitude. The data of the same mobile phone number is sorted by time to form a day's travel trajectory of the mobile phone user. At this time, the user's travel location is the location of the mobile phone base station. . Then, the cell phone base station is divided into the Thiessen polygon (the range of the cell phone base station is the service range of the cell phone base station), and multiple buildings located in the same Thiessen polygon as the cell phone base station are the candidate sets of individual positions.

A travel survey is an investigation of individual travel behavior. The travel survey data includes: individual work unit, departure place, destination, departure time, end time, travel mode, travel purpose, etc., but the location information is based on the traffic community. The travel trajectory of an individual without a mobile phone is constructed with each travel information in the travel survey, and multiple buildings in the same traffic area are obtained as a candidate set of individual locations.

The activity chain construction module 102 constructs an individual activity chain according to the candidate set of individual positions obtained above.

The contact network construction module 103 is used to dynamically construct a contact network of 24 time series in a day based on the activity chain of the constructed individual, with a time step of one hour. That is, the contact network construction module 103 dynamically constructs a contact network based on the individual activity chain of the urban population. According to the constructed activity chain of individuals, with the hour as the time granularity, compare the types of activities performed by different individuals within a day, and set individuals who perform the same activities at the same time and are in the same building location as individuals with the same time and space. Based on the activity category, different contact probabilities are assigned to individuals who co-occur in time and space. Under the constraint of contact probability, a dynamic contact network with hourly resolution between agents is generated. Wherein, the activity categories include: home, work/school, leisure and entertainment activities.

in particular:

An individual has 24 contact networks in a day. Take a contact network as an example. The vertices in the network represent individuals. Individuals appearing in the same position are connected by edges, which means that two individuals travel at the same location at the same time. Then, according to the contact probability in Table 1, it is judged whether two individuals appearing at the same place at the same time actually have contact. For example, when the infected person is an adult, he will have contact with a minor at a probability of 0.25 when he is at home. 0.4 probability of contact with an adult.

Table 1 Probability of contact between individuals

The spread simulation module 104 is used to simulate the spread of influenza at high temporal and spatial resolution by using the SEIR model according to the constructed contact network of 24 time series in a day. in particular:

The simulation process takes hours as the time unit to track when and where the infection occurred, and from whom it was transmitted to whom. During the spread of influenza, the probability of an individual being infected is called the effective probability of infection P. In the model, the formula for effective infection probability is as follows:

P=P _c ×P _i ×r

The analysis module 105 is used to analyze the simulation results from two perspectives of time and space, explore the trend and intensity of influenza outbreaks, analyze high-risk areas in the process of influenza transmission, and accurately locate key spatial locations in the process of influenza epidemic transmission.

Specifically:

The analysis module 105 performs a comparative analysis on the city scale and the administrative division scale from the time series. In order to accurately express the spread of influenza in time and space, the city is divided into 1km×1km grids, and the number of newly infected cases in each grid every day is analyzed. The number of cases is the average of 100 simulation results. The analysis module 105 constructs an epidemic tree by constructing the transmission topology relationship between the parent infected person and the offspring infected person, and then the epidemic forest is composed of multiple epidemic trees (Figure 3) to reveal the parent infected person and the offspring infected person The spatial transmission path between the two, accurately locates the key spatial locations in the spread of influenza epidemics, and reveals the relationship between the transmission distance and the transmission intensity of influenza in the city and the spatial agglomeration effect.

Example one of this application:

(1) Modeling of population attributes. According to the probability distribution of individual attributes such as age, gender, and occupation type in the census data, the corresponding individual attributes are assigned to each synthetic individual through Monte Carlo simulation. The age and gender distribution of the synthetic individual are shown in Figure 5. A synthetic family is constructed according to the probability of family attributes such as family category, family size, and family age components in the census data, and synthetic individuals are filled into the synthetic family. Figure 6 shows the size distribution of the synthetic family constructed in this embodiment. Compared with the population data of each district released by the 2017 Shenzhen Statistical Yearbook, the synthetic population constructed in this embodiment basically matches the population.

(2) At the overall level of the city, the simulated curve and the actual flu data are in a high degree of trend agreement, which can reflect important information such as the intensity of the flu outbreak, the duration of the outbreak, and the peak value. Figure 7 shows the comparison between the simulation results of the influenza spreading process and the real cases in time series. Figure 8 shows the number of regenerations generated by the model during the entire simulation cycle. The number of regenerations greater than 1 indicates that the epidemic is spreading, and less than 1 indicates the trend of the epidemic. To die. When the value of the regeneration number is greater than 1, the regeneration number first rises and then decreases, and the results of 100 simulations show that the value of the regeneration number fluctuates greatly in the part greater than 1. When the value of the regeneration number is less than 1, the regeneration number decreases slightly in the early stage and basically stabilizes in the later stage. The 100 simulation results show that the regeneration number is less volatile in the part less than 1.

(3) Under the administrative division scale, most administrative regions can still better reflect the intensity and trend of influenza outbreaks (Figure 9). Pingshan District, Longhua District, Guangming District, and Dapeng District have fewer hospitals and social health facilities, and the spatio-temporal density of influenza cases is sparse (the last row in Figure 9), resulting in large deviations in trends between the simulation results and the real influenza data.

(4) Figure 10 is the result of simulating the temporal and spatial distribution of influenza spread in Shenzhen. The value in each grid is the average of 100 simulation results. The agent-based spatial explicit infectious disease model constructed by fusing mobile phone location data proposed in this embodiment can support tracking the location information of infected persons and their interaction with other susceptible persons, effectively solving the inability to truly locate the space of infected persons. Location and potential high-risk transmission areas.

(5) Construct an epidemic tree based on the results of 100 simulations. Under the spatial scale of 1Km×1Km, this embodiment analyzes the importance of each spatial unit in the spread of the epidemic (Figure 11-14). Figure 11 reflects the number of influenza spreading within each spatial unit, that is, the homes of the parents of the infected and the offspring are located in the same grid, and such infections often occur between neighbors in the same community. It can be seen from Figure 11 that there is a high-risk area in the southeast of Luohu District. Influenza is more likely to spread infection within this location, which can remind residents living in this area to prevent the spread of influenza among neighbors. Figures 12, 13, and 14 reflect the situation where the infected parent and the infected offspring are in two spatial units during the spread of influenza. The linear propagation paths of the infected parent and the infected offspring of different spatial units constitute the connection relationship between the grids, and the connection relationship between the grids and the grids forms a social network. The value of the grid described in Figure 12 It is the degree of the grid in the social network, reflecting the number of grids that a grid can affect. It can be seen that the spatial unit at the junction of Futian District and Luohu District will have an impact on more geographic spaces in Shenzhen than other spatial units. Through analysis, it is found that when influenza spreads between grids, the weight of the connection relationship between the grids is different (the connection weight represents the number of events in which influenza transmission occurs between the two grids). The connection weight between 99.9% of the grids is less than 100 (Figure 13), and the area where the connection weight is greater than 100 has obvious spatial aggregation (Figure 14). Although the proportion of areas with a connection weight greater than 100 is very small, it is important to identify this part of the area, because this part of the area has a greater impact on the spread of influenza, the probability of influenza spread in this area is greater, and this part of the area has obvious space The phenomenon of aggregation provides convenience for the implementation of epidemic containment measures. The spatial distance between the grids represents the transmission distance of influenza, and the connection weight between the grids represents the transmission intensity of influenza between the grids. In order to reveal the relationship between the transmission distance and the transmission intensity, the Pearson between the two is calculated in this embodiment. The correlation coefficient r, r=-0.098, shows that the propagation distance and the propagation strength are extremely weakly negatively correlated.

The invention establishes a spatial explicit infectious disease model based on an agent. Based on the traditional method of individual movement modeling based on statistical data, a new spatio-temporal framework that integrates large-scale mobile phone location data for individual movement modeling is proposed, and the chain of individual urban activities is reconstructed. The model includes the entire population in the urban area (both demographic attributes and mobile behavior), the individual travel location is estimated to the building, and the individual activity place is based on the family (home), work unit (work), and building (entertainment). Construct a dynamic contact network between individuals. Then the SEIR model is used to simulate the time and space propagation process of influenza in the city. The simulation time step is 1 hour, the spatial scale is in the unit of building, and the heterogeneity of individual response to influenza virus is fully considered in the simulation process.

It should be noted that the present invention is not limited to the fusion of data such as mobile phone data, travel survey data, population census data, and building census data. The present invention can simulate the spread of a variety of infectious diseases, such as: influenza, dengue fever and other close-transmitted diseases. In addition, the present invention has good analysis capabilities for the spread of infectious diseases in time and space, especially spatial analysis. ability.

The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications are also It should be regarded as the protection scope of the present invention.

Claims

A method for simulating the spatiotemporal transmission process of influenza with large-scale trajectory data is characterized in that the method includes the following steps:

a. Synthesize urban population based on census data and building census data, and assign corresponding demographic attributes to individuals who synthesize urban population;

b. For individuals of synthetic urban population given demographic attributes, use mobile phone location data as the main and travel survey data as a supplement to construct an individual activity chain;

c. Based on the constructed individual activity chain, with one hour as the time step, dynamically construct a contact network of 24 time series in a day;

d. According to the constructed contact network of 24 time series in a day, the SEIR model is used to simulate the spread of influenza at high temporal and spatial resolution.
The method according to claim 1, wherein the method further comprises step e:

Analyze the simulation results from the two perspectives of time and space, obtain the spatial transmission path between the infected persons, and accurately locate the key spatial locations in the spread of the influenza epidemic.
The method of claim 1, wherein:

The population census data includes: age, gender, occupation type, family category, family size, and family age components; the building census data includes: building location information, building height, building area, and building function; wherein, the Building functions include: factories, teaching buildings, residential buildings, office buildings, shopping malls;

The demographic attributes include individual attributes, family attributes, and whether they are individuals with mobile phones; wherein, the individual attributes include age, gender, and occupation; the family attributes include family structure, home address, and work place.
The method according to claim 3, wherein said step b specifically includes the following steps:

The travel trajectory of an individual with a mobile phone is constructed based on the mobile phone location data: the data of the same mobile phone number is sorted by time to form the mobile phone user’s one-day travel trajectory, based on the mobile phone base station to divide the Tyson polygon, and the mobile phone base station is located in the same Tyson polygon The multiple buildings in are the candidate sets of individual locations;

Construct the travel trajectory of individuals without mobile phones based on travel survey data: construct each piece of travel information in the travel survey to obtain multiple buildings in the same traffic district as a candidate set of individual locations;

According to the candidate set of individual positions obtained above, an individual's activity chain is constructed.
The method according to claim 4, wherein said step c specifically comprises the following steps:

According to the activity chain of the constructed individuals, with the hour as the time granularity, compare the types of activities performed by different individuals in a day, and set individuals who perform the same activities at the same time and are in the same building location as individuals with co-occurrence in time and space, based on the activity category , To give different contact probabilities between individuals co-occurring in time and space. Under the constraint of contact probability, a dynamic contact network with hourly resolution between agents is generated.
The method according to claim 5, wherein the activity categories include: home, work/school, leisure and entertainment activities.
The method according to claim 5, wherein said step d specifically comprises the following steps:

With hours as the unit of time, track the time and place of the infection event, and from whom to whom. During the spread of influenza, the probability of an individual being infected is called the effective infection probability P. The formula for the effective infection probability is as follows:

P=P c ×P i ×r

Among them, P c is the contact probability between individuals, P i is the infection probability of the individual, and r is the relative infectivity of the individual;

Finally, the Monte Carlo method is used to determine whether the individual is infected.
8. The method of claim 7, wherein said determining whether the individual is infected by the Monte Carlo method comprises:

Based on a computer-generated uniformly distributed pseudo-random number, the pseudo-random number is compared with the effective probability of infection. If the pseudo-random number is less than or equal to the effective probability of infection, the individual is infected. Repeat the above process until the final trend of the number of new infections each day is consistent with the true The data is consistent and the calculated basic reproduction number is greater than 1.
The method according to claim 8, wherein said step e specifically comprises the following steps:

Conduct a comparative analysis on the time series at the city scale and the administrative division scale, divide the city into 1km×1km grids, and analyze the number of new infections in each grid every day;

The epidemic tree is constructed by constructing the spread topological relationship between the parent infected person and the offspring infected person, and then the epidemic forest is composed of multiple epidemic trees to obtain the spatial transmission path between the parent infected person and the offspring infected person, and accurately locate the flu The key spatial location in the spread of the epidemic.
A system for simulating the spatiotemporal transmission process of influenza with large-scale trajectory data, characterized in that the system includes a population attribute assignment module, an activity chain building module, a contact network building module, and a transmission simulation module, wherein:

The population attribute assignment module is used to synthesize urban population based on census data and building census data, and assign corresponding population attributes to individuals who synthesize urban population;

The activity chain building module is used to construct an individual activity chain based on mobile phone location data and travel survey data as a supplement to individuals of the synthetic urban population assigned population attributes;

The contact network building module is used to dynamically construct a contact network of 24 time series in a day based on the constructed individual activity chain, with one hour as a time step;

The spread simulation module is used to simulate the spread of influenza at high temporal and spatial resolution by using the SEIR model according to the constructed contact network of 24 time series in a day.