CN115130865A

CN115130865A - Risk identification method and device and electronic equipment

Info

Publication number: CN115130865A
Application number: CN202210764765.7A
Authority: CN
Inventors: 张娇娇
Original assignee: Hangzhou Dt Dream Technology Co Ltd
Current assignee: Hangzhou Dt Dream Technology Co Ltd
Priority date: 2022-06-29
Filing date: 2022-06-29
Publication date: 2022-09-30

Abstract

The embodiment of the application provides a risk identification method and device and electronic equipment. The method comprises the following steps: obtaining multidimensional user data from a plurality of different data sources; wherein the user data comprises data relating to user behavior; determining whether a target user matching a target population exists based on the multi-dimensional user data; and determining whether the target user has risk behaviors or not according to the user data corresponding to the target user.

Description

Risk identification method and device and electronic equipment

Technical Field

The embodiment of the application relates to the field of security, and in particular relates to a risk identification method and device, and electronic equipment.

Background

With the continuous promotion of city construction, the scale and population of each city are continuously increased, and the method also provides more problems about city management for city managers while realizing economic prosperity. For example, how to deal with a risk event involving security may occur anytime and anywhere.

In the related art, urban management relies heavily on video surveillance and is often the reason for tracing back through video surveillance after a risk event occurs. The early warning is difficult to give in time before the risk event happens, the coverage range of the video monitoring camera is limited, and a plurality of monitoring blind areas are easy to appear when the density of the camera is insufficient.

Disclosure of Invention

The embodiment of the specification provides a risk identification method and device and electronic equipment.

According to a first aspect of embodiments herein, there is provided a method of risk identification, the method comprising:

obtaining multidimensional user data from a plurality of different data sources; wherein the user data comprises data relating to user behavior;

determining whether a target user matching a target population exists based on the multi-dimensional user data;

and determining whether the target user has risk behaviors or not according to the user data corresponding to the target user.

According to a second aspect of embodiments herein, there is provided an apparatus for risk identification, the apparatus comprising:

the acquisition unit acquires multi-dimensional user data from a plurality of different data sources; wherein the user data comprises data relating to user behavior;

a determining unit that determines whether there is a target user matching a target population based on the multi-dimensional user data;

and the identification unit is used for determining whether the target user has risk behaviors or not according to the user data corresponding to the target user.

According to a third aspect of embodiments herein, there is provided an electronic apparatus including:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to perform any of the above risk identification methods.

The embodiment of the specification provides a risk identification scheme, and multi-dimensional user data provided by a plurality of different data sources are integrated to analyze target user behaviors so as to determine whether the target user has risk behaviors. On the one hand, due to the fact that multi-dimensional user data provided by various different data sources are integrated, the problem that risk identification is difficult to effectively conduct due to the fact that monitoring blind areas exist when single video monitoring is adopted can be avoided. On the other hand, whether the target user has the risk behavior can be predicted in advance through risk identification, and early warning is timely carried out when the risk behavior exists, so that the risk behavior of the target user can be interfered in advance, and the risk behavior is prevented from being converted into a risk event.

Drawings

FIG. 1 is a system architecture diagram for risk identification provided by an embodiment of the present description;

FIG. 2 is a flow diagram of a method of risk identification provided by an embodiment of the present description;

FIG. 3 is a hardware block diagram of a risk identification apparatus provided in an embodiment of the present specification;

fig. 4 is a schematic block diagram of a risk identification apparatus provided in an embodiment of the present specification.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the specification, as detailed in the claims that follow.

The terminology used in the description herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the description. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It should be understood that although the terms first, second, etc. may be used herein to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present description. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.

Referring now to fig. 1, a system architecture diagram for risk identification is shown, which may include a multidimensional data source 110, a server 120, a manager 130, and the like.

The multi-dimensional data source 110 may include a plurality of data sources with different dimensions, such as a video surveillance camera 111, a radio frequency identification reader 112, a communication device 113, and the like; also, the data source 110 may collect data related to user behavior and provide the data as user data to the server 120.

The server 120 may perform risk identification on the multidimensional user data provided by the multidimensional data source 110 to determine whether a risk behavior exists in the user. In addition, the server 120 may also push risk prompt information to the administrator 130 managing the user when recognizing that the user has risk behaviors.

After receiving the risk prompt information pushed by the server 120, the management party 130 may manage the user with the risk behavior, for example, refrain from continuing the risk behavior. In some embodiments, the server 120 may also skip the manager 130 to directly manage the user with the risk behavior.

When the method is implemented, the server side can synthesize multi-dimensional user data provided by various different data sources to analyze the target user behavior so as to determine whether the target user has risk behavior. On the one hand, due to the fact that multi-dimensional user data provided by various different data sources are integrated, the problem that risk identification is difficult to effectively conduct due to the fact that monitoring blind areas exist when single video monitoring is adopted can be avoided. On the other hand, whether the target user has the risk behavior can be predicted in advance through risk identification, and early warning is timely carried out when the risk behavior exists, so that the risk behavior of the target user can be interfered in advance, and the risk behavior is prevented from being converted into a risk event.

The following may be described by way of example with reference to a method for risk identification shown in fig. 2, which may be applied to the server in fig. 1, and which may include the following steps:

step 210: acquiring multidimensional user data from a plurality of different data sources; wherein the user data comprises data relating to user behavior.

Step 220: determining whether there is a target user matching the target population based on the multi-dimensional user data.

In practical applications, risk identification is usually performed on users belonging to a target group rather than on all users.

The target population generally refers to a user population of significant interest, for example, the target population includes a first type of population or a second type of population; wherein the first type of demographic comprises a managed demographic and the second type of demographic comprises a minor demographic.

In general, the managed population may include users who have criminal records; such users are typically managed in city management. The minors are easy to face danger due to the fact that the minors are small in age and not strong in self-protection consciousness; management of the minors group is also required in order to protect the minors.

In an exemplary embodiment, the multidimensional user data may include identity information of the user; accordingly, the step 220 may include:

matching the identity information of the user with an identity information base corresponding to a target crowd;

and if the matching is successful, determining the user corresponding to the identity information in the identity information base as a target user.

In this specification, whether a user belongs to a target group to be managed is determined by identity information of the user, and risk identification is further performed on the target user belonging to the target group. Since only risk identification needs to be performed on the target user, system resource overhead can be reduced.

In an exemplary embodiment, the identity information may include a facial image of the user or facial features extracted from the facial image; the identity information base comprises a face feature base;

the matching of the identity information of the user with the identity information base corresponding to the target crowd comprises:

and matching the facial features extracted from the facial image of the user with a facial feature library corresponding to the target crowd.

In this specification, the face image may be user data acquired by the camera for video monitoring, and the server determines the target user through a face recognition technology. Since the face image has uniqueness and the face image of each user is different, the target user is accurately determined based on the face image.

In another exemplary embodiment, the identity information may also include other information having a unique point to the user himself. Such as gait information; the gait information may refer to posture characteristic information of the user when walking. The gait information may also be user data collected by the video surveillance camera. Because the postures of different users are different when the users walk, the gait information of the users can be compared with the gait feature library, and the target users can be accurately determined.

The face feature library and the gait feature library can adopt a deep learning technology to extract features of check samples (face check images and posture check images) of users of a target group in advance, and construct corresponding feature libraries. Therefore, when the verification method is implemented, as long as the input user data is successfully matched with any feature in the feature library, the user corresponding to the input user data can be determined to belong to the target group and is the target user corresponding to the successfully matched verification sample.

Step 230: and determining whether the target user has risk behaviors or not according to the user data corresponding to the target user.

After the target user is determined, risk identification can be further performed by integrating the multidimensional user data corresponding to the target user to determine whether the target user has risk behaviors.

In an exemplary embodiment, the multi-dimensional user data may include a variety of data related to the geographic location of the user. For example, the data related to the geographic position of the user includes geographic position data of a camera where a picture of the user is collected, geographic position data of a Radio Frequency IDentification (RFID) reader that identifies a Radio Frequency IDentification (RFID) tag carried by the user, geographic position data of a communication device accessed to the user using the terminal, and the like.

In implementation, if a picture of a user appears in a video acquired by a camera, it indicates that the user has arrived at the position where the camera is located at that time, so that the geographical position data where the camera is located can be used as one user data of a target user.

Similarly, in some cases involving RFID, if the RFID reader recognizes the RFID tag carried by the user, it indicates that the user has arrived at the location of the RFID reader at that time, and therefore, the geographical location data of the RFID reader may be used as a kind of user data of the target user.

The same is true. If a communication device (e.g. a wireless base station, a wifi receiver, etc.) accesses a user terminal used by a user, it indicates that the user has arrived at the location of the communication device at the time, and therefore, the geographical location data of the communication device can be used as a kind of user data of a target user.

In the present specification, in the absence of geographic location data directly representing a user, geographic location data where the user is actually located may be indirectly determined by means of one or more data sources, so as to facilitate subsequent determination of whether the user has a risky behavior based on the geographic location data of the user.

After obtaining the various data related to the geographic location of the target user, step 230 may further include:

constructing an activity track of the target user according to a plurality of data related to the geographic position of the target user; the activity track is formed by serially connecting data related to geographic positions of users corresponding to different time points according to the sequence of the time points;

comparing the activity track of the target user with a historical conventional track of the target user or a geographic position area defined by a virtual electronic fence to determine whether the activity track of the target user is abnormal;

and if so, determining that the target user has risk behaviors.

In practical application, whether risk behaviors exist for a target user can be judged from the activity track of the target user, and if the activity track is greatly different from the historical conventional track, the risk behaviors may be indicated. The historical regular track can be generated by calculating a large number of historical tracks of the target person. The historical regular track represents an activity track that the target person often travels.

In practical applications, a management party usually sets a virtual electronic fence for managing a target group, and the target group is allowed to move only in a geographic location area defined by the virtual electronic fence or not allowed to move in the geographic location area defined by the virtual electronic fence. Once the target user triggers the virtual fence, it may indicate that a risky behavior exists.

The following exemplarily presents several application scenarios applicable to the embodiments of the present specification.

First introducing a management scenario for a target group of people, which may include a first type of people or a second type of people; wherein the first type of population comprises the aforementioned managed population and the second type of population comprises the aforementioned minors population.

In an exemplary embodiment, for a first type of crowd, the geographic location area bounded by the virtual electronic fence may comprise an area of permissible activity for the managed user;

correspondingly, the comparing the activity track of the target user with the geographic location area defined by the virtual electronic fence to determine whether the activity track of the target user is abnormal may include:

comparing the activity track of the managed target user with the geographic location area defined by the virtual electronic fence;

and if the activity track of the managed target user enters the geographic position area defined by the virtual electronic fence, determining that the activity track of the managed target user has an abnormality.

In this specification, through the virtual electronic fence technique, the risk identification of the managed crowd appearing in the geographic location area defined by the virtual electronic fence can be accurately realized. For example, for a user with heavy drinking, the risk early warning and timely intervention can be immediately carried out once the user appears in a place where alcohol can be purchased.

In an exemplary embodiment, for a second type of crowd, the geographic location area bounded by the virtual electronic fence includes a high risk area set for minors, which may include:

correspondingly, the comparing the activity track of the target user with the geographic location area defined by the virtual electronic fence to determine whether the activity track of the target user is abnormal includes:

comparing the target user's minor activity trajectory to the high risk areas delineated by the virtual electronic fence;

determining that the target user's activity trajectory for the underage is abnormal if the target user's activity trajectory enters a high risk area defined by the virtual electronic fence.

In the present specification, through the virtual electronic fence technology, the risk identification of the geographic location area defined by the virtual electronic fence in which the non-adult group appears can be accurately realized.

In an exemplary embodiment, the high risk zone may include a zone at risk of drowning.

As the minors belong to drowned high-risk groups, the description provides drowning risk early warning measures, once the minors enter an area with drowning risk, the risk early warning is triggered immediately, and related personnel are informed to dissuade the minors.

In an exemplary embodiment, the high risk areas may include places where entry is prohibited by minors; such as internet cafes, bars, etc. where entry by minors is prohibited.

The specification provides a sensitive place early warning measure, once a minor enters a sensitive place where the minor is prohibited from entering, risk early warning is triggered immediately, and related personnel are informed to dissuade the minor from adults.

In practical application, it is considered that minors generally belong to students, and the students have a fixed work and rest rule, such as learning in a campus in the daytime and resting in a student dormitory at night; thus, for a second type of population, the geographic location area bounded by the virtual electronic fence may also include a low risk area set for minors;

accordingly, the comparing the activity track of the target user with the geographic location area defined by the virtual electronic fence to determine whether the activity track of the target user is abnormal may include:

screening out an activity track in a preset time period from activity tracks of immature target users;

comparing the activity track in the preset time period with a low risk area defined by the virtual electronic fence;

and if the activity track in the preset time period leaves the low-risk area defined by the virtual electronic fence, determining that the activity track of the target user who is not an adult is abnormal.

In this specification, the preset time period may include a class time period, and the low risk area may include an area in a campus;

or, the preset time period comprises a sleeping time period, and the low risk area comprises a student dormitory.

When the method is implemented, the activity tracks of target users who are not adults in the class period are screened, and once the target users leave the campus in the class period, risk early warning is triggered immediately. And screening the activity tracks of target users who are not adults in the sleeping time period, and triggering risk early warning immediately once finding that the target users leave the student dormitory in the sleeping time period.

Through the embodiment, the behaviors of escaping from classes, staying at night and the like of students can be found in time, early warning and processing are carried out in time, and accidents of the students outside the school are avoided.

The above management scenarios for managed groups of people and minors are presented, followed by further description of the management scenarios for the areas of emphasis.

In real life, a management party may need to manage things in a focus area in addition to managing target people. For example, the management of illicit vehicles within a motor vehicle lane, the management of vehicles blocking fire paths, etc.

In an exemplary embodiment, the multidimensional user data includes vehicle driving data driven by a user;

accordingly, the step 230 may further include:

if the vehicle running data indicate that the vehicle is in a stop state, determining whether the vehicle driven by the target user has a risk behavior of violating a stop or blocking a fire fighting channel according to a parking position indicated by the vehicle running data;

and if the vehicle running data indicate that the vehicle is in a running state, determining whether the vehicle driven by the target user has the risk behavior of entering a forbidden area or not according to the running position indicated by the vehicle running data.

It is known that vehicle parking will seriously affect the normal traffic of other vehicles on the road, and if so, road congestion will be caused, and if so, traffic accidents may be caused; the fire fighting access is a life access for fire, and related regulations clearly indicate that the vehicle is prohibited from occupying the fire fighting access.

In the description, whether the dangerous behaviors exist or not is determined based on vehicle running data driven by a user, so that early warning on the dangerous behaviors of vehicle intrusion, vehicle illegal parking or fire channel blockage is realized.

In this specification, when it is determined that a target user has a risk behavior, the server may further push risk prompt information to a manager associated with the target user; and the management party intervenes in the target user, so that the risk event caused by the fact that the target user continues to carry out risk behaviors is avoided.

Therefore, the problem that risk identification is difficult to effectively carry out due to the existence of a monitoring blind area when single video monitoring is adopted can be avoided by integrating multi-dimensional user data provided by various different data sources; moreover, because the risk behaviors belong to potential risk events, the seedling heads of the risk events can be put out at the germination stage through prior intervention; in conclusion, the embodiment provided by the specification can effectively improve the treatment level and treatment effect of the city.

In practical application, the following embodiment can also perform risk identification on the acquired multidimensional user data by means of a machine learning technology, so that the identification speed and the identification accuracy of the risk identification can be improved.

In an exemplary embodiment, the foregoing step 220 may further include:

inputting a first type of user data in the multi-dimensional user data into a pre-constructed integrated model for first-stage calculation to determine whether a target user matched with a target crowd exists or not;

accordingly, the foregoing step 230 may further include:

under the condition that a target user matched with a target crowd exists, inputting a second type of user data in the multi-dimensional user data into the integration model for second-stage calculation to determine whether the target user has risk behaviors; wherein the ensemble model includes a model obtained by ensemble learning a plurality of machine learning models corresponding to the multidimensional user data

In this specification, a plurality of machine learning models corresponding to multidimensional user data may be used as weak learners, and one strong learner may be trained on the basis of the plurality of weak learners by an ensemble learning technique. This strong learner has a higher performance relative to each weak learner, enabling risks to be performed faster and more accurately.

In this specification, the first type of user data may refer to user data for determining whether there are target users matching a target population; the first type of user data may be multidimensional. Similarly, the second type of user data may refer to user data used to determine whether the target user has risky behavior; the second type of user data may also be multidimensional.

In this specification, for the correspondence between the multidimensional user data and the plurality of machine learning models, one machine learning model may be associated with each of the multidimensional user data.

For example, user data associated with a dimension of identity information may correspond to a machine learning model; the user data associated with the dimension of geographic location information may correspond to another machine learning model.

In addition, the corresponding relation between the multidimensional user data and the plurality of machine learning models can be set for fine granularity; for example, different user data in the same dimension may correspond to different machine learning models.

Taking the dimension of identity information as an example, if the user data includes face image information, gait information, and the like; then a machine learning model for face recognition (which may be generally referred to as a face recognition model) may be associated with the face image information, and a machine learning model for gait recognition (which may be generally referred to as a gait recognition model) may be associated with the similar face image information.

Because the user data with different dimensions have larger difference before, the calculation in one machine learning model is difficult, therefore, the user data with different dimensions are calculated by different machine learning models respectively, and the calculation complexity can be effectively reduced. In the same way, the user data under the same dimensionality may have larger difference, so that the single type of user data is calculated in a targeted manner through the machine learning model with a special function, the calculation complexity is reduced, and the calculation result is relatively more accurate and reliable because the machine learning model is specially used for calculating the type of user data.

In this specification, how many machine learning models or what kind of dimension corresponding machine learning models are specifically used may be flexibly configured according to actual needs.

In practical applications, a plurality of machine learning models corresponding to multidimensional user data are generally available; in this way, in the integrated learning process, as the weak learner is trained, the strong learner is further trained on the basis of a plurality of weak learners; without having to train a completely new model from scratch.

In an exemplary embodiment, the plurality of machine learning models includes a first machine learning model for performing a first stage computation and a second machine learning model for performing a second stage computation;

inputting a first type of user data in the multi-dimensional user data into a pre-constructed integrated model for a first stage of computation to determine whether a target user matching a target population exists, comprising:

respectively inputting the user data of each dimension in the first type of user data into a plurality of first machine learning models contained in the integrated model to obtain first user characteristics corresponding to the user data of each dimension; the first user characteristic is used for representing the identity attribute of the user from the corresponding dimension;

inputting the first user characteristic into a first classifier contained in the integration model to obtain a classification result aiming at the user;

determining that there is a target user matching a target population in a situation where the classification result indicates that the user is a target user;

in this specification, the ensemble learning model may be calculated in two stages, where the first stage is used to identify the target users who are matched with the target population, and the second stage is used to further identify whether the target users have risky behaviors after the target users are identified.

For example, it is assumed that the multidimensional user data obtained in step 210 and derived from a plurality of different data sources are: data A, data B, data C, data D, data E and data F; the data A, the data B and the data C belong to user data of a first type, and the data D, the data E and the data F belong to user data of a second type;

when the method is implemented, firstly, data A, data B and data C can be input into a plurality of first machine learning models included in the integrated model, for example, the data A is input into the first machine learning model A, the data B is input into the first machine learning model B, and the data C is input into the first machine learning model C; obtaining a first user characteristic A obtained by calculation of a first machine learning model A, a first user characteristic B obtained by calculation of a first machine learning model B and a first user characteristic C obtained by calculation of a first machine learning model C; then, inputting the first user characteristic A, the first user characteristic B and the first user characteristic C into a first classifier to obtain a classification result output by the first classifier and aiming at the user;

if the classification result for the user indicates that the user is a target user, further inputting data D, data E, and data F to a plurality of second machine learning models included in the ensemble model, such as data D to second machine learning model D, data E to second machine learning model E, and data F to second machine learning model F; obtaining a second user characteristic D obtained by calculation of a second machine learning model D, a second user characteristic E obtained by calculation of a second machine learning model E and a second user characteristic F obtained by calculation of a second machine learning model F; then, inputting a second user characteristic D, a second user characteristic E and a second user characteristic F into a second classifier to obtain a classification result output by the second classifier and aiming at the target user;

determining that the target user has the risk behavior if the classification result for the target user indicates that the risk behavior exists; on the contrary, if the classification result for the target user indicates that no risk behaviors exist, determining that no risk behaviors exist for the target user.

Corresponding to the method embodiment of risk identification, the present specification also provides an embodiment of a risk identification device. The device embodiments may be implemented by software, or by hardware, or by a combination of hardware and software. The software implementation is taken as an example, and as a logical device, the device is formed by reading corresponding computer business program instructions in the nonvolatile memory into the memory for operation through the processor of the device in which the device is located. In terms of hardware, as shown in fig. 3, a hardware structure diagram of a device where the apparatus for risk identification is located in this specification is shown, except for the processor, the network interface, the memory, and the nonvolatile memory shown in fig. 3, the device where the apparatus is located in the embodiment may also include other hardware according to the actual function of risk identification, which is not described again.

Referring to fig. 4, a block diagram of an apparatus for risk identification according to an embodiment of the present disclosure is provided, where the apparatus corresponds to the embodiment shown in fig. 2, and the apparatus includes:

an obtaining unit 410, obtaining multidimensional user data from a plurality of different data sources; wherein the user data comprises data relating to user behavior;

a determining unit 420, which determines whether there is a target user matching the target population based on the multi-dimensional user data;

the identifying unit 430 determines whether the target user has a risk behavior according to the user data corresponding to the target user.

In an exemplary embodiment, the multidimensional user data includes identity information of the user;

the determining unit 420 may further include:

the matching subunit matches the identity information of the user with an identity information base corresponding to the target crowd; and if the matching is successful, determining the user corresponding to the identity information in the identity information base as a target user.

In an exemplary embodiment, the identity information includes a face image of the user or a face feature extracted from the face image; the identity information base comprises a face feature base;

the matching subunit is further configured to match the face features extracted from the face image of the user with a face feature library corresponding to a target group.

In an exemplary embodiment, the multidimensional user data includes a plurality of data related to the geographic location of the user;

the identifying unit 430 may further include:

the construction subunit is used for constructing the activity track of the target user according to various data related to the geographical position of the target user; the activity track is formed by serially connecting data related to geographic positions of users corresponding to different time points according to the sequence of the time points;

the comparison subunit compares the activity track of the target user with a historical conventional track of the target user or a geographic position area defined by a virtual electronic fence to determine whether the activity track of the target user is abnormal; and if so, determining that the target user has risk behaviors.

In an exemplary embodiment, the data related to the geographic location of the user includes geographic location data where a camera acquiring a picture of the user is located, geographic location data where a radio frequency identification reader identifying a radio frequency identification tag carried by the user is located, and geographic location data where a communication device accessing the user terminal is located.

In an exemplary embodiment, the target population includes a first type of population or a second type of population; wherein the first type of demographic comprises a managed demographic and the second type of demographic comprises a minor demographic.

In an exemplary embodiment, for a first type of crowd, the geographic location area bounded by the virtual electronic fence comprises an area in which the managed user is allowed to be active;

the comparing subunit is further configured to compare the activity track of the managed target user with the geographic location area defined by the virtual electronic fence; and if the activity track of the managed target user enters the geographic position area defined by the virtual electronic fence, determining that the activity track of the managed target user has an abnormality.

In an exemplary embodiment, for a second type of crowd, the geo-location area bounded by the virtual electronic fence comprises a high risk area set for minors;

the comparing subunit is further operable to compare the activity track of the minor target user with the high risk area defined by the virtual electronic fence; determining that the target user's activity trajectory for the underage is abnormal if the target user's activity trajectory enters a high risk area defined by the virtual electronic fence.

In an exemplary embodiment, the high risk areas include areas at risk of drowning and/or places where entry is prohibited by minors.

In an exemplary embodiment, for a second type of population, the geographic location area bounded by the virtual electronic fence comprises a low risk area set for minors;

the comparison subunit is further configured to screen out an activity track within a preset time period from activity tracks of an immature target user; comparing the activity track in the preset time period with a low risk area defined by the virtual electronic fence; and if the activity track in the preset time period leaves the low-risk area defined by the virtual electronic fence, determining that the activity track of the target user who is not an adult is abnormal.

In an exemplary embodiment, the preset time period comprises a class time period, and the low risk area comprises an area in a campus;

the identification unit 430 is further configured to determine whether the vehicle driven by the target user has a risk behavior of violating parking or blocking a fire fighting tunnel according to a parking position indicated by the vehicle driving data if the vehicle driving data indicates that the vehicle is in a stopped state; and if the vehicle running data indicate that the vehicle is in a running state, determining whether the vehicle driven by the target user has the risk behavior of entering a forbidden area or not according to the running position indicated by the vehicle running data.

In an exemplary embodiment, the apparatus further comprises:

and the pushing unit is used for pushing risk prompt information to a manager associated with the target user when the target user is determined to have risk behaviors.

In an exemplary embodiment, the determining unit 420 may include:

and the first calculation subunit inputs the first type of user data in the multi-dimensional user data into a pre-constructed integrated model for first-stage calculation so as to determine whether a target user matched with the target crowd exists.

The identifying unit 430 may include:

the second calculation subunit is used for inputting second-stage calculation into the integrated model to determine whether the target user has risk behaviors or not under the condition that the target user matched with the target crowd exists; the integrated model comprises a model obtained by performing integrated learning on a plurality of machine learning models corresponding to the multidimensional user data.

the first computing subunit comprising: respectively inputting the user data of each dimension in the first type of user data into a plurality of first machine learning models contained in the integrated model to obtain first user characteristics corresponding to the user data of each dimension; the first user characteristic is used for representing the identity attribute of the user from the corresponding dimension; inputting the first user characteristic into a first classifier contained in the integration model to obtain a classification result aiming at the user; in the case of an indication for the user that the user is a target user, determining that there is a target user that matches a target demographic;

the second computing subunit includes: inputting the user data of each dimension in the second type of user data into a plurality of second machine learning models contained in the integrated model respectively to obtain second user characteristics corresponding to the user data of each dimension; the second user characteristic is used for representing the risk attribute of the user from the corresponding dimension; inputting the second user characteristics into a second classifier contained in the integrated model to obtain a classification result aiming at the target user; determining that the target user has the risk behavior in the case that the classification result for the target user indicates that the risk behavior exists.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. A typical implementation device is a computer, which may be in the form of a personal computer, laptop, cellular telephone, camera phone, smart phone, personal digital assistant, media player, navigation device, email messaging device, game console, tablet computer, wearable device, or a combination of any of these devices.

The implementation process of the functions and actions of each unit in the above device is specifically described in the implementation process of the corresponding step in the above method, and is not described herein again.

For the device embodiments, since they substantially correspond to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution in the specification. One of ordinary skill in the art can understand and implement it without inventive effort.

Fig. 4 above describes the internal functional modules and the structural schematic of the risk identification apparatus, and the substantial execution subject may be an electronic device, including:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to perform the method of risk identification of any of the preceding embodiments.

In the above embodiments of the electronic device, it should be understood that the Processor may be a Central Processing Unit (CPU), other general-purpose processors, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), etc. The general-purpose processor may be a microprocessor, or the processor may be any conventional processor, and the aforementioned memory may be a read-only memory (ROM), a Random Access Memory (RAM), a flash memory, a hard disk, or a solid state disk. The steps of a method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware processor, or may be implemented by a combination of hardware and software modules in a processor.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the embodiment of the electronic device, since it is substantially similar to the embodiment of the method, the description is simple, and for the relevant points, reference may be made to part of the description of the embodiment of the method.

Other embodiments of the present disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This specification is intended to cover any variations, uses, or adaptations of the specification following, in general, the principles of the specification and including such departures from the present disclosure as come within known or customary practice within the art to which the specification pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the specification being indicated by the following claims.

It will be understood that the present description is not limited to the precise arrangements described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present description is limited only by the appended claims.

Claims

1. A method of risk identification, the method comprising:

2. The method of claim 1, wherein the multidimensional user data includes identity information of a user;

the determining whether there is a target user matching a target population based on the multi-dimensional user data comprises:

3. The method according to claim 2, wherein the identity information comprises a face image of the user or a face feature extracted from the face image; the identity information base comprises a face feature base;

4. The method of claim 1, wherein the multidimensional user data comprises a plurality of data related to a geographic location of the user;

determining whether the target user has a risk behavior according to the user data corresponding to the target user includes:

and if so, determining that the target user has risk behaviors.

5. The method according to claim 4, wherein the data related to the geographical location of the user comprises the geographical location data of a camera which collects the picture of the user, the geographical location data of a radio frequency identification reader which identifies the radio frequency identification tag carried by the user, and the geographical location data of a communication device which accesses the user using the terminal.

6. The method of claim 4, wherein the target population comprises a first type of population or a second type of population; wherein the first type of demographic comprises a managed demographic and the second type of demographic comprises a minor demographic.

7. The method of claim 6, wherein for a first type of population, the geographic location area bounded by the virtual electronic fence comprises an area of permitted activity for the managed user;

the comparing the activity track of the target user with the geographic location area defined by the virtual electronic fence to determine whether the activity track of the target user is abnormal includes:

8. The method of claim 6, wherein for a second type of population, the geographic location area bounded by the virtual electronic fence comprises a high risk area set for minors;

comparing the activity track of the underage target user with the high risk area defined by the virtual electronic fence;

9. The method of claim 8, wherein the high risk areas include areas at risk of drowning and/or places where entry is prohibited by minors.

10. The method of claim 6, wherein for a second type of crowd, the geographic location area bounded by the virtual electronic fence comprises a low risk area set for minors;

11. The method of claim 10, wherein the preset time period comprises a class time period and the low risk area comprises an intra-campus area;

12. The method of claim 1, wherein the multidimensional user data includes vehicle travel data driven by the user;

13. The method of claim 1, further comprising:

and when determining that the target user has risk behaviors, pushing risk prompt information to a manager associated with the target user.

14. The method of claim 1, wherein determining whether there are target users matching a target demographic based on the multi-dimensional user data comprises:

under the condition that a target user matched with a target crowd exists, inputting a second type of user data in the multi-dimensional user data into the integration model for second-stage calculation to determine whether the target user has risk behaviors; the integrated model comprises a model obtained by performing integrated learning on a plurality of machine learning models corresponding to the multidimensional user data.

15. The method of claim 14, wherein the plurality of machine learning models comprises a first machine learning model for performing a first stage computation and a second machine learning model for performing a second stage computation;

in the case of an indication for the user that the user is a target user, determining that there is a target user that matches a target demographic;

inputting a second type of user data in the multidimensional user data into the integration model for second-stage calculation to determine whether the target user has a risk behavior, including:

inputting the user data of each dimension in the second type of user data into a plurality of second machine learning models contained in the integrated model respectively to obtain second user characteristics corresponding to the user data of each dimension; the second user characteristics are used for representing risk attributes of the target user from corresponding dimensions;

inputting the second user characteristics into a second classifier contained in the integrated model to obtain a classification result aiming at the target user;

determining that the target user has the risk behavior in the case that the classification result for the target user indicates that the risk behavior exists.

16. An apparatus for risk identification, the apparatus comprising:

the acquisition unit acquires multidimensional user data from a plurality of different data sources; wherein the user data comprises data relating to user behavior;

17. An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured as the method of any of the above claims 1-15.