Floating population identification walking access method and system

Abstract

The invention relates to a floating population identification walking access method, which comprises the following steps of: firstly, analyzing each to-be-analyzed residential address according to each primary analysis feature to find a suspicious address, and confirming in combination with a door-to-door registration mode, and then monitoring the change of the living floating population by each secondary analysis feature through a comparative analysis mode. In this way, periodic triggering execution is conducted on the whole process, the living floating population and changes of the living floating population are found in time in a monitoring mode, and the accuracy and timeliness of floating population registration are guaranteed; and a corresponding system is designed, the execution of the design method is efficiently realized and the accuracy in the execution process is ensured by matching the general scheduling subsystem with the mobile terminals equipped by the workers, so that the working efficiency of the design method and the system in practical application is comprehensively improved.

Description

Technical Field

[0001] The present invention relates to a floating population identification and visiting method and system, belonging to the floating population identification technical field. Background Art

[0002] Surveys of migrant populations are an important social research activity, with multidimensional impacts on social security, encompassing both the safety of migrants themselves and the potential risks they pose to social stability in the regions they migrate to, as well as measures to address these risks. Migrant population surveys directly impact social security by preventing and reducing crime, safeguarding public health, and preventing mass incidents. Indirectly, migrant population surveys can promote social integration and stability, and rationally allocate public resources. However, existing migrant population surveys face challenges with data accuracy, privacy protection, and dynamic tracking. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for identifying and interviewing floating population, designing explicit feature analysis and feature comparison analysis of floating population, realizing systematic periodic monitoring of floating population, and ensuring the accuracy and timeliness of floating population registration.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention designs a floating population identification and interview method, which performs the following steps for each residential address to be analyzed to realize floating population identification and interview;

[0005] According to the preset first cycle duration, the floating population identification method is executed according to the following steps A1 to A2;

[0006] Step A1. Based on the preset primary analysis features, the floating population identification model is applied to identify whether each residential address to be analyzed is a floating person. If so, the residential address to be analyzed of the floating person is updated to a suspicious address and the process proceeds to step A2; otherwise, no processing is performed;

[0007] Step A2. Based on the staff's door-to-door registration confirmation of each suspicious address, if it is confirmed that a migrant is living at the suspicious address, the suspicious address is defined as the target residential address. If it is confirmed that no migrant is living at the suspicious address, the suspicious address is restored to the residential address to be analyzed;

[0008] At the same time, according to the preset second cycle duration, the population change detection method is executed. The cycle is respectively for each target residential address, and the characteristic values of each secondary analysis feature preset in the preset second time period from the current time point to the historical time direction of the target residential address are collected, and the difference between each characteristic value and the characteristic value of the corresponding secondary analysis feature collected in the previous adjacent cycle is obtained. The population change identification model is applied to identify whether the population living in the target residential address has changed. If so, the target residential address is updated to the residential address to be analyzed, otherwise no processing is performed.

[0009] As a preferred technical solution of the present invention: before the execution of the floating population identification method and the change population detection method, a preset number of residential addresses corresponding to the confirmed floating person labels are initialized, and each residential address corresponds to the characteristic value of each first-level analysis feature preset within the preset first time period from its confirmation time point to the historical time direction, and the label corresponding to the residential address, combined with the initial timestamp, and the characteristic value of each first-level analysis feature corresponding to the residential address, constitute a first positive sample; at the same time, a preset number of residential addresses corresponding to the confirmed non-resident floating person labels are initialized, and each residential address corresponds to the characteristic value of each first-level analysis feature preset within the preset first time period from its confirmation time point to the historical time direction, and the label corresponding to the residential address, combined with the initial timestamp, and the characteristic value of each first-level analysis feature corresponding to the residential address, constitute a first negative sample; and then obtain each first positive sample and each first negative sample to form a first sample set;

[0010] Said step A2 further includes, for the target residential address of the residential migrant confirmed by the staff during on-site registration, constituting a residential address corresponding to the confirmed residential migrant label, and combining the current timestamp and the characteristic values of each first-level analysis feature obtained in step A1 corresponding to the residential address, constituting a first positive sample, and adding it to the first sample set; for the target residential address of the non-resident migrant confirmed by the staff during on-site registration, constituting a residential address corresponding to the confirmed non-resident migrant label, and combining the current timestamp and the characteristic values of each first-level analysis feature obtained in step A1 corresponding to the residential address, constituting a first negative sample, and adding it to the first sample set;

[0011] Follow steps i to v below to build a floating population identification model;

[0012] Step i. Determine whether the number of non-initialized first samples in the first sample set is greater than a preset new sample number threshold m. If so, first count the m first samples from the latest timestamp to the historical time direction, obtain the maximum eigenvalue and the minimum eigenvalue of each first-level analysis feature in each first positive sample in the m first samples, and form the fluctuation range of each first-level analysis feature in the standard first positive sample. At the same time, obtain the maximum eigenvalue and the minimum eigenvalue of each first-level analysis feature in each first negative sample in the m first samples, and form the fluctuation range of each first-level analysis feature in the standard first negative sample, and then proceed to step ii; otherwise, directly proceed to step iii;

[0013] Step ii. For each first sample in the first sample set other than the m first samples selected in step i, determine whether there are feature values in the first positive sample that do not meet the fluctuation range of each primary analysis feature in the standard first positive sample. If so, delete the first positive sample from the first sample set and update the first sample set; otherwise, do not process it; at the same time, determine whether there are feature values in the first negative sample that do not meet the fluctuation range of each primary analysis feature in the standard first negative sample. If so, delete the first negative sample from the first sample set and update the first sample set; otherwise, do not process it; after completing the determination and processing of the other first samples in the first sample set, proceed to step iii;

[0014] Step iii. Constructing a threshold comparison module for performing threshold comparison on the characteristic values of each primary analysis feature corresponding to the residential address based on the threshold parameters of each primary analysis feature, determining that the primary analysis feature that meets the threshold parameters corresponds to a label value of 0, and the primary analysis feature that does not meet the threshold parameters corresponds to a label value of 1, thereby obtaining a label vector consisting of the label values of each primary analysis feature corresponding to the residential address, and then proceeding to step iv;

[0015] Step iv. Based on the preset various neural network modules, each of which is used to implement the label vector corresponding to the residential address as input and the label corresponding to the residential address as output, the threshold comparison module connects each neural network module in the direction from the input end to the output end to construct each network to be trained, and then proceeds to step v;

[0016] Step v. For each network to be trained, based on each first sample in the first sample set, using the feature values of each primary analysis feature corresponding to the residential address in the first sample as input and the label corresponding to the residential address in the first sample as output, training the threshold parameters and parameters of the neural network module in the network to be trained, thereby obtaining the accuracy of each trained network and the accuracy of each trained network. The trained network with the highest accuracy is selected to form a migrant population identification model;

[0017] In step A1, the characteristic values of the first-level analysis features corresponding to each residential address to be analyzed within a preset first time period from the current time point to the historical time direction are first collected, and then the floating population identification model is applied to process each residential address to be analyzed respectively to identify whether there are floating people living in the residential address to be analyzed.

[0018] As a preferred technical solution of the present invention: according to the following steps I to II, a population change recognition model is constructed;

[0019] Step I. Based on a preset number of different combinations of individuals and the characteristic values of each preset secondary analysis feature corresponding to the residential address of each combination of individuals within a preset second time period, a second negative sample is constructed using two different combinations of individuals and the characteristic values of each secondary analysis feature corresponding to the second time period, and the difference between the characteristic values of each secondary analysis feature between the two combinations of individuals in the second negative sample is obtained, and the label of the residential population change corresponding to the second negative sample is obtained, thereby obtaining each second negative sample;

[0020] At the same time, based on each person combination and the characteristic values of each secondary analysis feature preset in the two preset second time periods corresponding to the residential address of each person combination, a second positive sample is constructed using the characteristic values of each secondary analysis feature of a single person combination and its corresponding two second time periods, and the difference between the characteristic values of each secondary analysis feature between the two second time periods in the second positive sample is obtained, and the label of the residential population corresponding to the second positive sample has not changed, thereby obtaining each second positive sample;

[0021] Then proceed to step II;

[0022] Step II. Based on each second negative sample and each second positive sample, with each difference in the second sample as input and the label corresponding to the second sample as output, a preset neural network is trained to obtain a population change recognition model.

[0023] As a preferred technical solution of the present invention: the first-level analysis features include residential stability, social activity density, nighttime electricity consumption dispersion, and community abnormal records; the second-level analysis features include social activity density and nighttime electricity consumption dispersion;

[0024] Among them, residential stability is based on the continuous payment records of water, electricity, gas and other energy sources corresponding to the residential address within the preset first period, according to the following formula:

[0025]

[0026] Calculate and obtain the characteristic value S of residential stability corresponding to the first preset time period of the residential address, where I=3, t1 represents the number of consecutive payment months of water energy corresponding to the first preset time period of the residential address, t2 represents the number of consecutive payment months of electricity energy corresponding to the first preset time period of the residential address, t3 represents the number of consecutive payment months of gas energy corresponding to the first preset time period of the residential address, T max Indicates the maximum number of consecutive payment months for water, electricity, and gas energy corresponding to the residential address within the preset first period; w1, w2, and w3 represent the preset weights corresponding to water, electricity, and gas energy respectively;

[0027] The density of social activities is calculated based on the number of express deliveries and takeout orders within the first and second preset time periods corresponding to the residential address, according to the following formula:

[0028] D = log 10 (1+0.5P+0.3F)

[0029] Calculate the characteristic value D of the social activity density within the corresponding length of time corresponding to the residential address, where P represents the average monthly number of express deliveries within the corresponding length of time corresponding to the residential address, and F represents the average monthly number of takeout orders within the corresponding length of time corresponding to the residential address;

[0030] The nighttime electricity consumption dispersion is based on the nighttime electricity consumption in the first and second preset time periods corresponding to the residential address, according to the following formula:

[0031]

[0032] Calculate the characteristic value V of the nighttime electricity consumption dispersion within the corresponding time period of the residential address, where N represents the number of days within the corresponding time period, and x n Indicates the early morning electricity consumption of the nth day within the corresponding time period corresponding to the residential address. Indicates the average electricity consumption in the early morning of each day within the corresponding time period corresponding to the residential address;

[0033] Community abnormal records are based on the records of residential addresses during the community visit within the corresponding preset first period, according to the following formula:

[0034] Score=∑ t∈T TF(t)×IDF(t)×100

[0035] Calculate the characteristic value Score of the abnormal records of the community within the first time period corresponding to the residential address, where T represents the preset keyword set related to the floating population, and TF(t) represents the number of keywords belonging to T in the record documents of the residential address during the community visit within the preset first time period. K represents the total number of records of residential addresses during the community visit within the preset first time period, df tIt indicates the number of record documents containing the keyword T in the record documents of residential addresses during the community visit within the preset first time period.

[0036] As an optimal technical solution of the present invention: for each target residential address of the confirmed mobile residents, the staff also conducts door-to-door registration according to a preset third cycle duration, and the third cycle duration is longer than the second cycle duration.

[0037] As a preferred technical solution of the present invention: based on executing the floating population identification method according to a preset first cycle duration, if the ratio of the cumulative number of suspicious addresses restored as residential addresses to be analyzed in step A2 to the cumulative number of suspicious addresses obtained in step A1 is greater than a preset threshold value after the preset number of consecutive cycle operations, then steps i to v are triggered to perform training and update the floating population identification model.

[0038] As a preferred technical solution of the present invention: regarding the staff's home visit, the record includes the residential address information and the staff's photo, and the staff's home visit registration includes registering the resident's identity information, communication method, and SMS verification of the communication method.

[0039] As a preferred technical solution of the present invention: in step A2, regarding the staff's visit to the suspicious address, if no one is at the suspicious address, the staff will be arranged to visit the address a second time, and the time between the two visits will be less than the time of the first cycle. If no one is at the address during both visits, the suspicious address will be directly defined as the target residential address of the migrant workers.

[0040] Corresponding to the above, the technical problem that the present invention also needs to solve is to provide a system for implementing a method for visiting and registering migrant population identification. Through a modular design approach, the design method can be efficiently implemented to improve the work efficiency of visiting and registering migrant population.

[0041] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention designs a system for a method of identifying and visiting migrant population, comprising a control module, various mobile terminals, and a data acquisition module, a data preprocessing module, a model training module, and a model application module respectively connected to the control module;

[0042] The control module realizes the forwarding of data between the data acquisition module, data preprocessing module, model training module, and model application module;

[0043] The data acquisition module is used to collect the characteristic time series values of each first-level analysis feature within a preset first time period corresponding to the residential address to be analyzed, and to collect the characteristic time series values of each second-level analysis feature within a preset second time period corresponding to the target residential address;

[0044] The data preprocessing module is used to preprocess the characteristic time series values corresponding to each primary analysis feature of the residential address to be analyzed collected by the data acquisition module to obtain the characteristic values of each primary analysis feature corresponding to the residential address to be analyzed, and first preprocess the characteristic time series values corresponding to each secondary analysis feature of the target residential address to obtain the characteristic values of each secondary analysis feature corresponding to the target residential address, and then process to obtain the characteristic value difference of each secondary analysis feature between two adjacent periods corresponding to the target residential address;

[0045] The model training module is used to build, train, and update the floating population identification model; and to train the population change identification model;

[0046] The model application module is used to apply the floating population identification model, process the characteristic values of each primary analysis feature corresponding to the residential address to be analyzed obtained by the data preprocessing module, and identify whether there are floating people living in the residential address to be analyzed; and to apply the population change identification model, process the characteristic value difference of each secondary analysis feature corresponding to the target residential address obtained by the data preprocessing module, and identify whether there is a change in the population living in the target residential address;

[0047] Each mobile terminal is equipped by each staff member, and each mobile terminal is respectively connected to the control module. Based on the staff's door-to-door visit, the mobile terminal is used to register the residents and upload the registration information to the control module.

[0048] As a preferred technical solution of the present invention: the structures of the mobile terminals are the same, and each mobile terminal includes an information entry module, an image acquisition module, and a text message verification module; wherein the information entry module is used to enable the staff to enter the identity information of the residents and upload it to the control module; the image acquisition module is used to collect photos containing residential address information and staff, and upload them to the control module to provide evidence of the staff's visit; the text message verification module is used to perform text message verification of the residents' communication methods.

[0049] The method and system for identifying and visiting migrant populations described in the present invention, using the above technical solution, has the following technical effects compared with the existing technology:

[0050] (1) The present invention designs a method for visiting and identifying floating population. First, the first-level analysis features are used to analyze each residential address to be analyzed, suspicious addresses are found, and confirmed by door-to-door registration. Then, the second-level analysis features are used to monitor changes in the floating population through comparative analysis. In this way, the entire process is triggered and executed periodically. In a monitoring manner, the floating population and changes in the floating population are discovered in a timely manner, thereby ensuring the accuracy and timeliness of the floating population registration. A corresponding system is designed, which uses the general dispatching subsystem and the mobile terminals equipped by each staff member to efficiently implement the execution of the design method and ensure the accuracy of the execution process, thereby comprehensively improving the work efficiency of the design method and system in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the architecture application of the floating population identification and visiting method and system designed by the present invention. DETAILED DESCRIPTION

[0052] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0053] The present invention designs a method and system for identifying and visiting floating population. In practical applications, such as Figure 1 As shown, first, the first-level analysis features are defined, including residential stability, social activity density, nighttime electricity consumption dispersion, and community abnormal records, and the second-level analysis features are defined, including social activity density and nighttime electricity consumption dispersion.

[0054] The design residential stability is based on the continuous payment records of water, electricity, gas and other energy sources corresponding to the residential address within the preset first period, according to the following formula:

[0055]

[0056] Calculate and obtain the characteristic value S of residential stability corresponding to the first preset time period of the residential address, where I=3, t1 represents the number of consecutive payment months of water energy corresponding to the first preset time period of the residential address, t2 represents the number of consecutive payment months of electricity energy corresponding to the first preset time period of the residential address, t3 represents the number of consecutive payment months of gas energy corresponding to the first preset time period of the residential address, T max It represents the maximum number of consecutive payment months for water, electricity and gas energy within the preset first period corresponding to the residential address; w1, w2 and w3 represent the weights corresponding to the preset water, electricity and gas energy respectively.

[0057] The social activity density is designed based on the number of express deliveries and takeouts within the first and second preset time periods corresponding to the residential address, according to the following formula:

[0058] D = log10 (1+0.5P+0.3F)

[0059] The characteristic value D of the social activity density corresponding to the residential address in the corresponding period is calculated, where P represents the average monthly number of express deliveries in the corresponding period, and F represents the average monthly number of takeouts in the corresponding period.

[0060] The design nighttime electricity consumption dispersion is based on the nighttime electricity consumption in the preset first time period and the preset second time period corresponding to the residential address, according to the following formula:

[0061]

[0062] Calculate the characteristic value V of the nighttime electricity consumption dispersion within the corresponding time period of the residential address, where N represents the number of days within the corresponding time period, and x n Indicates the early morning electricity consumption of the nth day within the corresponding time period corresponding to the residential address. It indicates the average electricity consumption in the early morning of each day within the corresponding time period of the residential address.

[0063] The design of community abnormality records is based on the record documents of residential addresses during the community visit within the corresponding preset first time period, according to the following formula:

[0064] Score=∑ t∈T TF(t)×IDF(t)×100

[0065] Calculate the characteristic value Score of the abnormal records of the community within the first time period corresponding to the residential address, where T represents the preset keyword set related to the floating population, and TF(t) represents the number of keywords belonging to T in the record documents of the residential address during the community visit within the preset first time period. K represents the total number of records of residential addresses during the community visit within the preset first time period, df t It indicates the number of record documents containing the keyword T in the record documents of residential addresses during the community visit within the preset first time period.

[0066] In actual design, such as Figure 1 As shown, the design system includes a general scheduling subsystem and mobile terminals equipped for each staff member. The general scheduling subsystem includes a control module, and a data acquisition module, a data preprocessing module, a model training module, and a model application module respectively connected to the control module. Each mobile terminal is respectively communicated with the control module.

[0067] Based on the design and definition of the above-mentioned first-level analysis features and second-level analysis features, a method for identifying and visiting migrant population is further designed. In actual application, the control module is used to realize the forwarding of data among the data acquisition module, data preprocessing module, model training module, and model application module. For each residential address to be analyzed, the following two schemes are executed simultaneously to realize the identification and visiting of migrant population.

[0068] One option, such as Figure 1 As shown, according to the preset first cycle length, the floating population identification method is executed according to the following steps A1 to A2.

[0069] Step A1. First, the data acquisition module connects to the public utility database (water, electricity, gas), accesses the community grid data, and retrieves the express / takeout data to collect the characteristic time series values of each first-level analysis feature corresponding to each residential address to be analyzed within the preset first time length from the current time point to the historical time direction. The data preprocessing module performs preprocessing to obtain the characteristic values of each first-level analysis feature corresponding to each residential address to be analyzed.

[0070] The model application module then applies the floating population identification model to process the characteristic values of each first-level analysis feature corresponding to each residential address to be analyzed, and identifies whether there are floating persons living in the residential address to be analyzed. If so, the residential address to be analyzed of the floating persons is updated to a suspicious address and enters step A2; otherwise, no processing is performed.

[0071] Step A2. Based on the staff's use of mobile terminals to confirm the door-to-door registration of each suspicious address, if it is confirmed that a migrant worker lives at the suspicious address, the suspicious address is defined as the target residential address. If it is confirmed that no migrant worker lives at the suspicious address, the suspicious address is restored to the residential address to be analyzed.

[0072] In actual application, in the above step A2, regarding the staff's visit to the suspicious address, if no one is at the suspicious address, the staff will be arranged to visit the address a second time, and the time between the two visits will be less than the time of the first cycle. If no one is found at the address during both visits, the suspicious address will be directly defined as the target residential address of the migrant worker.

[0073] Execute the second plan at the same time, such as Figure 1As shown, according to the preset second cycle length, the population change detection method is executed, and the cycle is respectively for each target residential address. First, the data acquisition module collects the characteristic time series values of each secondary analysis feature preset within the preset second time length from the current time point to the historical time direction corresponding to the target residential address, and the data preprocessing module performs preprocessing to obtain the characteristic values of each secondary analysis feature corresponding to the target residential address, and then further processes to obtain the difference between each characteristic value and the characteristic value of the corresponding secondary analysis feature collected in the adjacent previous cycle, that is, the characteristic value difference of each secondary analysis feature between the two adjacent cycles corresponding to the target residential address.

[0074] The model application module then applies the population change identification model to process the characteristic value differences of the target residential address corresponding to each secondary analysis feature to identify whether the population living in the target residential address has changed. If so, the target residential address is updated to the residential address to be analyzed, otherwise no processing is performed.

[0075] In actual application, regarding the acquisition of the floating population identification model involved in one of the above-mentioned schemes, before the floating population identification method and the change population detection method are executed, a preset number of residential addresses corresponding to the confirmed floating person labels, and each residential address corresponds to the characteristic value of each first-level analysis feature preset within the preset first time period from its confirmation time point to the historical time direction, and the label corresponding to the residential address, combined with the initial timestamp, and the characteristic value of each first-level analysis feature corresponding to the residential address, constitute a first positive sample; at the same time, a preset number of residential addresses corresponding to the confirmed non-resident floating person labels are initialized, and each residential address corresponds to the characteristic value of each first-level analysis feature preset within the preset first time period from its confirmation time point to the historical time direction, and the label corresponding to the residential address, combined with the initial timestamp, and the characteristic value of each first-level analysis feature corresponding to the residential address, constitute a first negative sample; and then obtain each first positive sample and each first negative sample to form a first sample set.

[0076] Said step A2 also includes: for the target residential address of the residential migrant personnel confirmed by the staff during door-to-door registration, a residential address corresponding to the confirmed residential migrant label is formed, and the first positive sample is formed in combination with the current timestamp and the characteristic values of the first-level analysis features obtained in step A1 corresponding to the residential address, and the first positive sample is added to the first sample set; for the target residential address of the non-resident migrant personnel confirmed by the staff during door-to-door registration, a residential address corresponding to the confirmed non-resident migrant label is formed, and the first negative sample is formed in combination with the current timestamp and the characteristic values of the first-level analysis features obtained in step A1 corresponding to the residential address, and the first negative sample is added to the first sample set.

[0077] Based on the construction of the first sample set, the model training module constructs a floating population recognition model according to the following steps i to v.

[0078] Step i. Determine whether the number of non-initialized first samples in the first sample set is greater than a preset new sample number threshold m. If so, first count the m first samples from the latest timestamp to the historical time direction, obtain the maximum eigenvalue and the minimum eigenvalue of each first-level analysis feature in each first positive sample in the m first samples, and constitute the fluctuation range of each first-level analysis feature in the standard first positive sample. At the same time, obtain the maximum eigenvalue and the minimum eigenvalue of each first-level analysis feature in each first negative sample in the m first samples, and constitute the fluctuation range of each first-level analysis feature in the standard first negative sample, and then enter step ii; otherwise, directly enter step iii.

[0079] Step ii. For each first sample in the first sample set other than the m first samples selected in step i, determine whether there is a feature value in the first positive sample that does not meet the fluctuation range of each first-level analysis feature in the standard first positive sample. If so, delete the first positive sample from the first sample set and update the first sample set; otherwise, do not process it; at the same time, determine whether there is a feature value in the first negative sample that does not meet the fluctuation range of each first-level analysis feature in the standard first negative sample. If so, delete the first negative sample from the first sample set and update the first sample set; otherwise, do not process it; after completing the determination and processing of each other first sample in the first sample set, proceed to step iii.

[0080] Step iii. Construct a threshold comparison module for performing threshold comparison on the feature values of each first-level analysis feature corresponding to the residential address based on the threshold parameters of each first-level analysis feature, and determine that the first-level analysis feature that meets the threshold parameters corresponds to a label value of 0, and the first-level analysis feature that does not meet the threshold parameters corresponds to a label value of 1, thereby obtaining a label vector composed of the label values of each first-level analysis feature corresponding to the residential address, and then proceeding to step iv.

[0081] Step iv. Based on the preset various neural network modules, each of which is used to implement the label vector corresponding to the residential address as input and the label corresponding to the residential address as output. From the input end to the output end, the threshold comparison module connects each neural network module respectively to construct each network to be trained, and then enters step v.

[0082] Step v. For each network to be trained, based on each first sample in the first sample set, with the feature values of each first-level analysis feature corresponding to the residential address in the first sample as input and the label corresponding to the residential address in the first sample as output, train each threshold parameter in the network to be trained and the parameters in the neural network module, and then obtain each trained network and the accuracy corresponding to each trained network, select the trained network corresponding to the highest accuracy, and form a floating population identification model.

[0083] In actual application, for the above-mentioned migrant population identification model that has been trained and used for actual implementation, the migrant population identification method is executed according to the preset first cycle duration. If the ratio of the cumulative number of suspicious addresses restored as residential addresses to be analyzed through step A2 to the cumulative number of suspicious addresses obtained through step A1 is greater than the preset threshold value after the preset number of cycles, the model training module is triggered to execute steps i to v to train and update the migrant population identification model.

[0084] In actual application, the population change identification model involved in one of the above solutions is constructed by the model training module according to the following steps I to II.

[0085] Step I. Based on a preset number of different personnel combinations and the characteristic values of each secondary analysis feature preset within a preset second time period corresponding to the residential address of each personnel combination, a second negative sample is constructed using two different personnel combinations and their corresponding characteristic values of each secondary analysis feature within the second time period, and the difference between the characteristic values of each secondary analysis feature between the two personnel combinations in the second negative sample is obtained, and the label of the residential population corresponding to the second negative sample has changed, thereby obtaining each second negative sample.

[0086] At the same time, based on each personnel combination and the characteristic values of each secondary analysis feature preset in the two preset second time periods corresponding to the residential address of each personnel combination, the second positive sample is constructed with the characteristic values of each secondary analysis feature of a single personnel combination and its corresponding two second time periods, and the difference between the characteristic values of each secondary analysis feature between the two second time periods in the second positive sample is obtained, and the label of the residential population corresponding to the second positive sample has not changed, thereby obtaining each second positive sample.

[0087] Then proceed to step II.

[0088] Step II. Based on each second negative sample and each second positive sample, with each difference in the second sample as input and the label corresponding to the second sample as output, a preset neural network is trained to obtain a population change recognition model.

[0089] While executing the above-mentioned design scheme one, step A1 to step A2, and the scheme two, respectively, for confirming the target residential addresses of the mobile residents, it also includes the staff conducting door-to-door registration according to the preset third cycle duration, and the third cycle duration is longer than the second cycle duration.

[0090] During the actual implementation of the design scheme of the present invention, the staff involved in the door-to-door visit are designed to use mobile terminals to record the residential address information and photos of the staff, and upload them to the control module in the general dispatch subsystem to provide evidence of the staff's door-to-door visit. The staff also use mobile terminals to register the identity information of the residents and upload it to the control module, and use mobile terminals to send text messages to the residents' communication methods to perform text message verification on the residents' communication methods.

[0091] The above technical solution designs a visiting method for identifying floating population. First, each first-level analysis feature is used to analyze each residential address to be analyzed, suspicious addresses are found, and confirmed by door-to-door registration. Then, each second-level analysis feature is used to monitor changes in the floating population through comparative analysis. In this way, the entire process is triggered and executed periodically. In a monitoring manner, the floating population and changes in the floating population are discovered in a timely manner, thereby ensuring the accuracy and timeliness of the floating population registration. A corresponding system is designed to efficiently implement the execution of the design method and ensure the accuracy of the execution process by using the overall dispatching subsystem and the mobile terminals equipped by each staff member. In this way, the work efficiency of the design method and system in actual application is comprehensively improved.

[0092] The embodiments of the present invention are described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made within the scope of knowledge possessed by ordinary technicians in this field without departing from the spirit of the present invention.

Claims

1. A method for identifying and interviewing floating population, characterized by: For each residential address to be analyzed, perform the following steps to identify and interview the floating population; According to the preset first cycle duration, the floating population identification method is executed according to the following steps A1 to A2; Step A1. Based on the preset primary analysis features, the floating population identification model is applied to identify whether each residential address to be analyzed is a floating person. If so, the residential address to be analyzed of the floating person is updated to a suspicious address and the process proceeds to step A2; otherwise, no processing is performed; Step A2. Based on the staff's door-to-door registration confirmation of each suspicious address, if it is confirmed that a migrant is living at the suspicious address, the suspicious address is defined as the target residential address. If it is confirmed that no migrant is living at the suspicious address, the suspicious address is restored to the residential address to be analyzed; At the same time, according to the preset second cycle duration, the population change detection method is executed. The cycle is respectively for each target residential address, and the characteristic values of each secondary analysis feature preset in the preset second time period from the current time point to the historical time direction of the target residential address are collected, and the difference between each characteristic value and the characteristic value of the corresponding secondary analysis feature collected in the previous adjacent cycle is obtained. The population change identification model is applied to identify whether the population living in the target residential address has changed. If so, the target residential address is updated to the residential address to be analyzed, otherwise no processing is performed.

2. A method for identifying and interviewing floating population according to claim 1, characterized in that: Before the floating population identification method and the change population detection method are executed, a preset number of residential addresses corresponding to the confirmed floating person labels are initialized, and the characteristic values of the first-level analysis features corresponding to each residential address within the preset first time period from the confirmation time point to the historical time direction are respectively corresponded, and the labels corresponding to the residential addresses, combined with the initial timestamp, and the characteristic values of the first-level analysis features corresponding to the residential addresses, constitute a first positive sample; at the same time, a preset number of residential addresses corresponding to the confirmed non-resident floating person labels are initialized, and the characteristic values of the first-level analysis features corresponding to each residential address within the preset first time period from the confirmation time point to the historical time direction are respectively corresponded, and the labels corresponding to the residential addresses, combined with the initial timestamp, and the characteristic values of the first-level analysis features corresponding to the residential addresses, constitute a first negative sample; and then each first positive sample and each first negative sample are obtained to form a first sample set; Said step A2 further includes, for the target residential address of the residential migrant confirmed by the staff during on-site registration, constituting a residential address corresponding to the confirmed residential migrant label, and combining the current timestamp and the characteristic values of each first-level analysis feature obtained in step A1 corresponding to the residential address, constituting a first positive sample, and adding it to the first sample set; for the target residential address of the non-resident migrant confirmed by the staff during on-site registration, constituting a residential address corresponding to the confirmed non-resident migrant label, and combining the current timestamp and the characteristic values of each first-level analysis feature obtained in step A1 corresponding to the residential address, constituting a first negative sample, and adding it to the first sample set; Follow steps i to v below to build a floating population identification model; Step i. Determine whether the number of non-initialized first samples in the first sample set is greater than a preset new sample number threshold m. If so, first count the m first samples from the latest timestamp to the historical time direction, obtain the maximum eigenvalue and the minimum eigenvalue of each first-level analysis feature in each first positive sample in the m first samples, and form the fluctuation range of each first-level analysis feature in the standard first positive sample. At the same time, obtain the maximum eigenvalue and the minimum eigenvalue of each first-level analysis feature in each first negative sample in the m first samples, and form the fluctuation range of each first-level analysis feature in the standard first negative sample, and then proceed to step ii; Otherwise go directly to step iii; Step ii. For each first sample in the first sample set other than the m first samples selected in step i, determine whether there are feature values in the first positive sample that do not meet the fluctuation range of each primary analysis feature in the standard first positive sample. If so, delete the first positive sample from the first sample set and update the first sample set; otherwise, do not process it; at the same time, determine whether there are feature values in the first negative sample that do not meet the fluctuation range of each primary analysis feature in the standard first negative sample. If so, delete the first negative sample from the first sample set and update the first sample set; otherwise, do not process it; after completing the determination and processing of the other first samples in the first sample set, proceed to step iii; Step iii. Constructing a threshold comparison module for performing threshold comparison on the characteristic values of each primary analysis feature corresponding to the residential address based on the threshold parameters of each primary analysis feature, determining that the primary analysis feature that meets the threshold parameters corresponds to a label value of 0, and the primary analysis feature that does not meet the threshold parameters corresponds to a label value of 1, thereby obtaining a label vector consisting of the label values of each primary analysis feature corresponding to the residential address, and then proceeding to step iv; Step iv. Based on the preset various neural network modules, each of which is used to implement the label vector corresponding to the residential address as input and the label corresponding to the residential address as output, the threshold comparison module connects each neural network module in the direction from the input end to the output end to construct each network to be trained, and then proceeds to step v; Step v. For each network to be trained, based on each first sample in the first sample set, using the feature values of each primary analysis feature corresponding to the residential address in the first sample as input and the label corresponding to the residential address in the first sample as output, training the threshold parameters and parameters of the neural network module in the network to be trained, thereby obtaining the accuracy of each trained network and the accuracy of each trained network. The trained network with the highest accuracy is selected to form a migrant population identification model; In step A1, the characteristic values of the first-level analysis features corresponding to each residential address to be analyzed within a preset first time period from the current time point to the historical time direction are first collected, and then the floating population identification model is applied to process each residential address to be analyzed respectively to identify whether there are floating people living in the residential address to be analyzed.

3. The method for identifying and visiting floating population according to claim 1, characterized in that: Follow steps I to II below to construct a population change identification model; Step I. Based on a preset number of different combinations of individuals and the characteristic values of each preset secondary analysis feature corresponding to the residential address of each combination of individuals within a preset second time period, a second negative sample is constructed using two different combinations of individuals and the characteristic values of each secondary analysis feature corresponding to the second time period, and the difference between the characteristic values of each secondary analysis feature between the two combinations of individuals in the second negative sample is obtained, and the label of the residential population change corresponding to the second negative sample is obtained, thereby obtaining each second negative sample; At the same time, based on each person combination and the characteristic values of each secondary analysis feature preset in the two preset second time periods corresponding to the residential address of each person combination, a second positive sample is constructed using the characteristic values of each secondary analysis feature of a single person combination and its corresponding two second time periods, and the difference between the characteristic values of each secondary analysis feature between the two second time periods in the second positive sample is obtained, and the label of the residential population corresponding to the second positive sample has not changed, thereby obtaining each second positive sample; Then proceed to step II; Step II. Based on each second negative sample and each second positive sample, with each difference in the second sample as input and the label corresponding to the second sample as output, a preset neural network is trained to obtain a population change recognition model.

4. A method for identifying and visiting floating population according to claim 2 or 3, characterized in that: The first-level analysis features include residential stability, social activity density, nighttime electricity consumption dispersion, and community abnormal records; the second-level analysis features include social activity density and nighttime electricity consumption dispersion; Among them, residential stability is based on the continuous payment records of water, electricity, gas and other energy sources corresponding to the residential address within the preset first period, according to the following formula: Calculate and obtain the characteristic value S of residential stability corresponding to the first preset time period of the residential address, where I=3, t1 represents the number of consecutive payment months of water energy corresponding to the first preset time period of the residential address, t2 represents the number of consecutive payment months of electricity energy corresponding to the first preset time period of the residential address, t3 represents the number of consecutive payment months of gas energy corresponding to the first preset time period of the residential address, T max Indicates the maximum number of consecutive payment months for water, electricity, and gas energy corresponding to the residential address within the preset first period; w1, w2, and w3 represent the preset weights corresponding to water, electricity, and gas energy respectively; The density of social activities is calculated based on the number of express deliveries and takeout orders within the first and second preset time periods corresponding to the residential address, according to the following formula: D=log 10 1+0.5P+0.3F) Calculate the characteristic value D of the social activity density within the corresponding length of time corresponding to the residential address, where P represents the average monthly number of express deliveries within the corresponding length of time corresponding to the residential address, and F represents the average monthly number of takeout orders within the corresponding length of time corresponding to the residential address; The nighttime electricity consumption dispersion is based on the nighttime electricity consumption in the first and second preset time periods corresponding to the residential address, according to the following formula: Calculate the characteristic value V of the nighttime electricity consumption dispersion within the corresponding time period of the residential address, where N represents the number of days within the corresponding time period, and x n Indicates the early morning electricity consumption of the nth day within the corresponding time period corresponding to the residential address. Indicates the average electricity consumption in the early morning of each day within the corresponding time period corresponding to the residential address; Community abnormal records are based on the records of residential addresses during the community visit within the corresponding preset first period, according to the following formula: Score=∑ t∈T TF(t)×IDF(t)×100 Calculate the characteristic value Score of the abnormal records of the community within the first time period corresponding to the residential address, where T represents the preset keyword set related to the floating population, and TF(t) represents the number of keywords belonging to T in the record documents of the residential address during the community visit within the preset first time period. K represents the total number of records of residential addresses during the community visit within the preset first time period, df t It indicates the number of record documents containing the keyword T in the record documents of residential addresses during the community visit within the preset first time period.

5. The method for identifying and interviewing floating population according to claim 1, characterized in that: For each target residential address of confirmed migrant workers, the staff will conduct door-to-door registration according to the preset third cycle duration, and the third cycle duration is longer than the second cycle duration.

6. The method for identifying and visiting floating population according to claim 2, characterized in that: Based on executing the floating population identification method according to the preset first cycle duration, if the ratio of the cumulative number of suspicious addresses restored as residential addresses to be analyzed in step A2 to the cumulative number of suspicious addresses obtained in step A1 is greater than the preset threshold value after the preset number of consecutive cycle operations, then steps i to v are triggered to perform training and update the floating population identification model.

7. A method for identifying and visiting floating population according to claim 1 or 5, characterized in that: Regarding the staff's home visit, the records include the residential address information and the staff's photo, and the staff's home registration includes the registration of the resident's identity information, communication method, and SMS verification of the communication method.

8. A method for identifying and visiting floating population according to claim 1 or 5, characterized in that: Regarding the staff's visit to the suspicious address in step A2, if no one is at the suspicious address, the staff will be arranged to visit the address a second time, and the time between the two visits is less than the time of the first cycle. If no one is at the address during both visits, the suspicious address will be directly defined as the target residential address of the migrant workers.

9. A system for implementing the method for identifying and visiting migrant population according to any one of claims 1 to 6, characterized in that: It includes a control module, various mobile terminals, and a data acquisition module, a data preprocessing module, a model training module, and a model application module respectively connected to the control module; The control module realizes the forwarding of data between the data acquisition module, data preprocessing module, model training module, and model application module; The data acquisition module is used to collect the characteristic time series values of each first-level analysis feature within a preset first time period corresponding to the residential address to be analyzed, and to collect the characteristic time series values of each second-level analysis feature within a preset second time period corresponding to the target residential address; The data preprocessing module is used to preprocess the characteristic time series values corresponding to each primary analysis feature of the residential address to be analyzed collected by the data acquisition module to obtain the characteristic values of each primary analysis feature corresponding to the residential address to be analyzed, and first preprocess the characteristic time series values corresponding to each secondary analysis feature of the target residential address to obtain the characteristic values of each secondary analysis feature corresponding to the target residential address, and then process to obtain the characteristic value difference of each secondary analysis feature between two adjacent periods corresponding to the target residential address; The model training module is used to build, train, and update the floating population identification model; and to train the population change identification model; The model application module is used to apply the floating population identification model, process the characteristic values of each primary analysis feature corresponding to the residential address to be analyzed obtained by the data preprocessing module, and identify whether there are floating people living in the residential address to be analyzed; and to apply the population change identification model, process the characteristic value difference of each secondary analysis feature corresponding to the target residential address obtained by the data preprocessing module, and identify whether there is a change in the population living in the target residential address; Each mobile terminal is equipped by each staff member, and each mobile terminal is respectively connected to the control module. Based on the staff's door-to-door visit, the mobile terminal is used to register the residents and upload the registration information to the control module.

10. The system for the mobile population identification and interviewing method according to claim 9, characterized in that: The structures of the mobile terminals are the same, and each mobile terminal includes an information entry module, an image acquisition module, and a text message verification module; among them, the information entry module is used to enable staff to enter the identity information of residents and upload it to the control module; the image acquisition module is used to collect photos containing residential address information and staff, and upload them to the control module to provide evidence of staff visits; the text message verification module is used for residents to perform text message verification of communication methods.

Images