CN115905624B

CN115905624B - Method, device and equipment for determining user behavior state

Info

Publication number: CN115905624B
Application number: CN202211271412.XA
Authority: CN
Inventors: 吕乐; 傅幸; 王维强; 张长浩; 筴硕; 林晓彤
Original assignee: Alipay Hangzhou Information Technology Co Ltd
Current assignee: Alipay Hangzhou Information Technology Co Ltd
Priority date: 2022-10-18
Filing date: 2022-10-18
Publication date: 2023-06-16
Anticipated expiration: 2042-10-18
Also published as: CN115905624A

Abstract

The embodiment of the specification discloses a method, a device and equipment for determining a user behavior state, wherein the method comprises the following steps: acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, determining a reconstruction coefficient corresponding to the service time sequence information based on the service time sequence information and a pre-trained neural network model, and optimizing model parameters in the neural network model in the training process of the neural network model through the following objective functions: an objective function determined based on norms constructed from the training samples and the reconstructed coefficient samples corresponding to the training samples, and entropy corresponding to the reconstructed coefficient samples; determining the association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information; and determining behavior state information of the target user for executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

Description

Method, device and equipment for determining user behavior state

Technical Field

The present document relates to the field of computer technologies, and in particular, to a method, an apparatus, and a device for determining a user behavior state.

Background

In many business scenarios, a user may perform a plurality of different operations, such as login, identity authentication, consumption, etc., and in particular, in payment business scenarios, the user may perform operations such as login, transfer, consumption, etc., where the operations may further include a large amount of attribute information, such as transfer amount, shopping category, etc., and the behavior of the user may be abstracted into a large amount of high-dimensional information that varies with time, so as to form a time sequence, which may be referred to as business timing information.

To describe a behavior pattern of a user performing a certain service or user account, it is often necessary to divide its service timing information into sub-sequence information, each representing a state of the user performing a certain service behavior pattern or representing a state of the user account behavior pattern. The behavior of the user can be considered to be continuous and stable in a behavior state in the subsequence information, the behavior of the user or the behavior of the user account can be considered to be mutated between different subsequence information, the stability in the subsequence information is helpful to describe the safety of the user account or the user to execute a certain service, and the state mutation among the subsequence information is helpful to characterize whether the user account is at risk or whether the user to execute a certain service is at risk. Although the behavior state information of the user account is critical, the behavior state information is usually missing in a practical scenario. Based on this, for the situation of lacking calibration, a technical scheme for determining different user behavior states more quickly and effectively needs to be provided.

Disclosure of Invention

The embodiment of the specification aims to provide a technical scheme for determining different user behavior states more quickly and effectively.

In order to achieve the above technical solution, the embodiments of the present specification are implemented as follows:

the embodiment of the specification provides a method for determining a user behavior state, which comprises the following steps: acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, and the attribute information of the target service which is sequentially arranged according to the time sequence comprises the operation behavior information generated by the target user executing the target service. Based on the service time sequence information and a pre-trained neural network model, determining a reconstruction coefficient corresponding to the service time sequence information, and optimizing model parameters in the neural network model in the process of training the neural network model through the following objective functions: and determining an objective function based on norms constructed by the training samples and the reconstruction coefficient samples corresponding to the training samples and entropy corresponding to the reconstruction coefficient samples. And determining the association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information. And determining behavior state information of the target user executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

The embodiment of the specification provides a device for determining a user behavior state, which comprises: the time sequence acquisition module acquires service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, and the attribute information of the target service which is sequentially arranged according to the time sequence comprises the operation behavior information generated by the target user executing the target service. The reconstruction coefficient determining module is used for determining a reconstruction coefficient corresponding to the service time sequence information based on the service time sequence information and a pre-trained neural network model, and optimizing model parameters in the neural network model through the following objective functions in the process of training the neural network model: and determining an objective function based on norms constructed by the training samples and the reconstruction coefficient samples corresponding to the training samples and entropy corresponding to the reconstruction coefficient samples. And the association relation determining module is used for determining association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information. And the behavior state determining module is used for determining the behavior state information of the target user executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

The embodiment of the specification provides a user behavior state determining device, which comprises: a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to: acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, and the attribute information of the target service which is sequentially arranged according to the time sequence comprises the operation behavior information generated by the target user executing the target service. Based on the service time sequence information and a pre-trained neural network model, determining a reconstruction coefficient corresponding to the service time sequence information, and optimizing model parameters in the neural network model in the process of training the neural network model through the following objective functions: and determining an objective function based on norms constructed by the training samples and the reconstruction coefficient samples corresponding to the training samples and entropy corresponding to the reconstruction coefficient samples. And determining the association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information. And determining behavior state information of the target user executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

The present description also provides a storage medium for storing computer-executable instructions that when executed by a processor implement the following: acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, and the attribute information of the target service which is sequentially arranged according to the time sequence comprises the operation behavior information generated by the target user executing the target service. Based on the service time sequence information and a pre-trained neural network model, determining a reconstruction coefficient corresponding to the service time sequence information, and optimizing model parameters in the neural network model in the process of training the neural network model through the following objective functions: and determining an objective function based on norms constructed by the training samples and the reconstruction coefficient samples corresponding to the training samples and entropy corresponding to the reconstruction coefficient samples. And determining the association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information. And determining behavior state information of the target user executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

Drawings

In order to more clearly illustrate the embodiments of the present description or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some of the embodiments described in the present description, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a diagram illustrating an embodiment of a method for determining a behavior state of a user according to the present disclosure;

FIG. 2 is a diagram illustrating another embodiment of a method for determining a behavior state of a user according to the present disclosure;

FIG. 3 is a schematic diagram of a graph segmentation process according to the present disclosure;

FIG. 4 is a schematic diagram of another process for segmenting a map according to the present disclosure;

FIG. 5 is a diagram illustrating another embodiment of a method for determining a behavior state of a user according to the present disclosure;

FIG. 6 is a diagram illustrating an embodiment of a user behavior state determination apparatus according to the present disclosure;

fig. 7 is an embodiment of a user behavior state determining apparatus according to the present specification.

Detailed Description

The embodiment of the specification provides a method, a device and equipment for determining a user behavior state.

In order to make the technical solutions in the present specification better understood by those skilled in the art, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only some embodiments of the present specification, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.

Example 1

As shown in fig. 1, the embodiment of the present disclosure provides a method for determining a user behavior state, where an execution subject of the method may be a terminal device or a server, where the terminal device may be a certain terminal device such as a mobile phone, a tablet computer, or a computer device such as a notebook computer or a desktop computer, or may also be an IoT device (specifically, such as a smart watch, an in-vehicle device, or the like). The server may be a single server, a server cluster including a plurality of servers, a background server such as a financial service or an online shopping service, or a background server of an application program. In this embodiment, a server is taken as an example for detailed description, and the following related contents may be referred to for the execution process of the terminal device, which is not described herein. The method specifically comprises the following steps:

In step S102, service timing information constructed by operation behavior information generated by the target user executing the target service for multiple times is obtained, where the service timing information includes information of the target service sequentially arranged according to a time sequence in which the target user executes the target service, and the attribute information of the target service sequentially arranged according to the time sequence includes operation behavior information generated by the target user executing the target service.

The target user may be any user, and in this embodiment, the target user may be a user who performs the target service. The target service may include various kinds, for example, a payment service, a transfer service, a face recognition service, etc., and may be specifically set according to actual conditions, which is not limited in the embodiment of the present specification. The operation behavior information may be information generated by any operation performed by the target user during the process of executing the target service by the target user, and the operation behavior information may specifically include payment amount, transfer amount, shopping category, and the like, and may specifically be set according to actual situations.

In practice, in many business scenarios, the user may perform various operations, such as login, identity authentication, consumption, etc., and in particular, in payment business scenarios, the user may perform operations such as login, transfer, consumption, etc., where the operations may further include a large amount of attribute information, such as transfer amount, shopping category, etc., and the behavior of the user may be abstracted into a large amount of high-dimensional information that varies with time, so as to form a time sequence, which may be referred to as business timing information. To describe a behavior pattern of a user performing a certain service or user account, it is often necessary to divide its service timing information into sub-sequence information, each representing a state of the user performing a certain service behavior pattern or representing a state of the user account behavior pattern. The behavior of the user can be considered to be continuous and stable in a behavior state in the subsequence information, the behavior of the user or the behavior of the user account can be considered to be mutated between different subsequence information, the stability in the subsequence information is helpful to describe the safety of the user account or the user to execute a certain service, and the state mutation among the subsequence information is helpful to characterize whether the user account is at risk or whether the user to execute a certain service is at risk. For example, a certain enterprise registers a payment account, the enterprise may operate multiple services such as catering, commodity retail, and the like, and the collection behavior patterns of different service scenes are obviously different, so that different behavior states in the service time sequence information under mixed operation need to be mined. As another example, some business-registered payment accounts have no action for a period of time, but only pay their employees at the end of each month, with silent, alternate patterns of behavior enabled, and as such, need to be reasonably described. Although the behavior state information of the user account is critical, the behavior state information is usually missing in a practical scenario.

Generally, to solve industries with high-dimensional informationThe problem of division of the traffic sequence information can be realized by using a subspace clustering algorithm, and can be realized by using

Represents a piece of service time sequence information, wherein the time sequence information comprises N time steps, namely the number of times that a user executes a certain service is N times, or the user executes a certain service at different time points of N times, and the like, x is _i The user performs attribute information (such as payment amount, payment time, location information, etc.) corresponding to a service for the i-th time, wherein the total of D dimensions of information is that the service time sequence information of the user behavior generally has self-expression property (self-expression property), that is, the attribute information corresponding to the service performed once by the user can be reconstructed by performing the attribute information corresponding to the service at other times of the user, and the property satisfies the assumption of sub-sequence information distribution, so that the service time sequence information can be segmented by using a subspace clustering algorithm.

The division of the traffic timing information can be performed by the following expression,

wherein, the liquid crystal display device comprises a liquid crystal display device,

z _i is to reconstruct X using the vector in X _i In which case the reconstruction coefficients may be linear combination coefficients, i.e. using +.>

Estimating x _i Wherein diag (Z) =0, i.e. Z _ii =0, expressed in reconstruction x _i In the process (1), only X is divided by X _i Information of a certain service executed at other time points. R in the above expression is a matrix of N× (N-1) which is the form:

ZR＝[z ₂ -z ₁ ，z ₃ -z ₂ ，…，z _N -z _N-1 ]

||ZR|| _1，2 the purpose of (a) is to force adjacent vectors in Z to be as similar as possible, since event x can be assumed in general among the traffic timing information _i And x _i-1 With consistency, the attributes of adjacent events should be similar. Lambda (lambda) ₁ And lambda (lambda) ₂ Parameters are respectively.

By solving the expression, a corresponding Z is obtained, and then the similarity relation between the high-dimensional information at each time point in the sequence X can be obtained. The regularization term Z is ₁ Will result in a final solution z _i The vectors are as sparse as possible, i.e. as few events x as possible _j Reconstruction x _i . If x _j Reconstruction x _i Coefficient z of (2) _ij Not 0, at the same time

The reconstruction loss of (a) is small, x can be considered as _j And x _i Has a certain similarity relationship. Can take w=z+z ^T Then element W in W _ij Describe x _i And x _j The information of each time point in the service time sequence information forms a graph structure in the mode, and the graph structure can be segmented by using a normalization segmentation algorithm of the graph, so that the segmentation of the service time sequence information can be realized.

However, in the above manner, for the service timing information X of each user account, the correlation relationship between each time point of the user account needs to be obtained by solving the above expression, so that the calculation complexity of the above process is high, and it is difficult to accelerate the above calculation process by using a GPU or other devices, based on which, for the situation of lacking calibration, a technical scheme for determining different user behavior states more quickly and effectively needs to be provided. The embodiment of the specification provides a technical scheme which can be realized, and the technical scheme specifically comprises the following steps:

For a target service, target users each timeWhen executing the target service, the relevant information of the target user for executing the target service can be recorded, wherein the recorded information can comprise the time for executing the target service, the information generated in the process of executing the target service, the position information and the like, for example, when the target user executes the target service for the first time, the information can be recorded as x ₁ ，x ₁ The attribute information of (2) may include the time of first executing the target service, the information generated during the first executing the target service, the position information, etc., and the second time the target user executes the target service may be recorded as x ₂ ，x ₂ The attribute information of (2) may include the time of executing the target service for the second time, information generated during the second time of executing the target service, position information, etc., and may be recorded as x when the target user executes the target service for the third time ₃ ，x ₃ The attribute information of (2) may include the time of executing the target service for the third time, information generated during the execution of the target service for the third time, position information, etc., and may be recorded as x when the target user of … … executes the target service for the nth time _N ，x _N The attribute information of (a) may include time of nth execution of the target service, information generated during nth execution of the target service, location information, etc. Based on the above, the target services can be sequentially arranged according to the time sequence of the target user executing the target services, namely

Wherein, the target business x is arranged in sequence according to time sequence ₁ ，x ₂ ，…，x _N The attribute information of (a) includes operation behavior information generated when the target user executes the target service, and the operation behavior information may include information of D dimensions, where the operation behavior information may include, for example, time of executing the target service each time, information generated during executing the target service each time, location information, and the like, and may be specifically set according to actual situations, which is not limited in the embodiment of the present specification.

By the method, the service time sequence information constructed by the operation behavior information generated by the target user executing the target service for many times can be obtained, and when the behavior state information of the target user executing the target service needs to be determined, the service time sequence information constructed by the operation behavior information generated by the target user executing the target service for many times can be obtained.

In step S104, based on the service timing information and the pre-trained neural network model, a reconstruction coefficient corresponding to the service timing information is determined, and in the process of training the neural network model, model parameters in the neural network model are optimized by the following objective functions: an objective function is determined based on norms constructed from the training samples and the reconstructed coefficient samples corresponding to the training samples, and the entropy corresponding to the reconstructed coefficient samples.

The neural network model may include various types, such as a convolutional neural network model, a cyclic neural network model, and the like, and may be specifically set according to actual situations, which is not limited in the embodiment of the present disclosure. The reconstruction coefficient corresponding to the service time sequence information can be as in the above

Can use +.>

Estimating x _i 。

In practice, a corresponding algorithm may be obtained, and a neural network model may be constructed based on the algorithm, where input data of the neural network model may be service timing information constructed for operation behavior information generated by a user performing a service multiple times, output data may be a reconstruction coefficient corresponding to the input data, and then a training sample for training the neural network model (i.e., service timing information constructed for operation behavior information generated by a user performing a service multiple times) may be obtained, and the training sample may be used to model-train the neural network model, where in the process of model training, considering the time sequence x in the training sample under the actual service scenario, which is simple ₁ ，x ₂ ，…，x _N The training samples are reconstructed by linear combination, which is unreasonable, if the reconstruction coefficient has negative numbers, reasonable interpretation is lacking in the actual service scene, and for this purpose, an objective function can be preset, and the method can be used for The model parameters in the neural network model are optimized based on the objective function, wherein the expression can be adjusted according to the objective function, namely, the regularization term Z can be adjusted ₁ The remaining terms may remain unchanged, i.e. the objective function is determined based on the norms constructed from the training samples and the reconstructed coefficient samples corresponding to the training samples, and the entropy corresponding to the reconstructed coefficient samples, adjusted to the entropy corresponding to the reconstructed coefficients. And then, training the neural network model by using a training sample, and simultaneously optimizing the model parameters through the objective function to finally obtain the trained neural network model.

After the service time sequence information is obtained in the mode, the service time sequence information can be input into the pre-trained neural network model, and corresponding output data, namely a reconstruction coefficient corresponding to the service time sequence information, can be obtained through processing of the neural network model.

In step S106, based on the reconstruction coefficient corresponding to the service timing information and the service timing information, an association relationship between the target services executed at different times in the service timing information is determined.

In implementation, after the reconstruction coefficient corresponding to the service time sequence information is obtained in the above manner, the target service executed at different times in the service time sequence information can be analyzed based on the reconstruction coefficient, so as to obtain the association relationship between the target services executed at different times in the service time sequence information.

In step S108, behavior state information of the target user for executing the target service is determined based on the association relationship between the target services executed at different times in the service timing information.

In implementation, the association relationship between the target services executed at different times in the service time sequence information can be analyzed, the relationship between the target service executed at any one time in the service time sequence information and the target service executed at other times can be determined from the association relationship, and then the target services of similar or same category in the relationship can be determined to be aggregated together, so that one or more different aggregation clusters can be obtained, each obtained aggregation cluster can be used as behavior state information, and the behavior state information of the target user executing the target service can be obtained based on the manner.

The embodiment of the specification provides a method for determining a user behavior state, which is implemented by acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, attribute information of the target service which is sequentially arranged according to the time sequence comprises operation behavior information generated by the target user executing the target service, then, based on the service time sequence information and a pre-trained neural network model, determining a reconstruction coefficient corresponding to the service time sequence information, and optimizing model parameters in the neural network model through the following target functions in the process of training the neural network model: based on the norm constructed by the training samples and the reconstructed coefficient samples corresponding to the training samples and the target function determined by the entropy corresponding to the reconstructed coefficient samples, then, the association relationship between the target services executed at different times in the service time sequence information can be determined based on the reconstructed coefficient corresponding to the service time sequence information and the service time sequence information, finally, the behavior state information of the target user executing the target service is determined based on the association relationship between the target services executed at different times in the service time sequence information, thus, the reconstructed coefficient corresponding to the service time sequence information is determined based on the pre-trained neural network model, the correlation relationship of the data in the service time sequence information is further extracted and clustering processing is carried out, the neural network model can output probability weight (namely, the reconstructed coefficient corresponding to the service time sequence information) as an encoder, the calculation process of the neural network model is faster than that of the optimization algorithm based on the sparse coding extracted data correlation relationship, and the hardware acceleration processing is easy to use GPU and the like, so that different behavior states of different users can be determined more quickly and effectively under the condition of lack of calibration, and the behavior state of different users can be determined.

Example two

As shown in fig. 2, the embodiment of the present disclosure provides a method for determining a user behavior state, where an execution subject of the method may be a terminal device or a server, where the terminal device may be a certain terminal device such as a mobile phone, a tablet computer, or a computer device such as a notebook computer or a desktop computer, or may also be an IoT device (specifically, such as a smart watch, an in-vehicle device, or the like). The server may be a single server, a server cluster including a plurality of servers, a background server such as a financial service or an online shopping service, or a background server of an application program. In this embodiment, a server is taken as an example for detailed description, and the following related contents may be referred to for the execution process of the terminal device, which is not described herein. The method specifically comprises the following steps:

in step S202, a training sample for training a neural network model is obtained, where the training sample is historical service timing information constructed by historical operation behavior information generated by a user executing a target service for multiple times, the historical service timing information includes information of the target service sequentially arranged according to a time sequence of the user executing the target service, and attribute information of the target service sequentially arranged according to the time sequence includes historical operation behavior information generated by the user executing the target service.

The neural network model comprises one or more of a cyclic neural network model, a convolutional neural network model and a neural network model based on an attention mechanism, and in practical application, the neural network model can be the model, can also comprise a plurality of different models, and can be specifically set according to practical situations.

In step S204, model training is performed on the neural network model based on the training sample, and model parameters in the neural network model are optimized through a gradient back propagation algorithm, so that the model parameters satisfy an objective function, and a trained neural network model is obtained.

Wherein the objective function is based on the training sample sumThe norm of the reconstruction coefficient sample construction corresponding to the training sample and the entropy determination corresponding to the reconstruction coefficient sample are used for simply carrying out time sequence x in the training sample under the actual service scene ₁ ，x ₂ ，…，x _N The process of reconstructing training samples by linear combination is not reasonable, if the reconstruction coefficient has negative numbers, reasonable interpretation is lacking in the actual service scene, and the reconstruction coefficient is used for reconstructing the time sequence x in the training samples ₁ ，x ₂ ，…，x _N Mixing is performed, and furthermore, it is necessary to ensure that the reconstruction coefficients are non-negative and that the sum of the reconstruction coefficients is 1. If the sum of the reconstruction coefficients is a fixed value, the regularization term Z in the above expression ₁ Cannot be applied. Regular term Z ₁ The role in the neural network model is to ensure that target services executed at other time points as few as possible are used in the process of reconstructing training samples by adding sparse regularization, so that the regularization term Z can be used for reconstructing training samples ₁ The entropy of the reconstruction coefficient is adjusted, the smaller the entropy, the more concentrated the reconstruction coefficient, the higher the certainty, that is, the target service executed at a given time point can be reconstructed by using the target service executed at a single or a small number of other time points, and the remaining terms can be kept unchanged (i.e., the third term λ in the objective function described below ₄ ||ZR|| _1，2 Is a canonical term for characterizing continuity, so the third term does not need to be adjusted), i.e., the objective function is as follows:

wherein lambda is ₃ And lambda (lambda) ₄ Parameters, Θ represents model parameters of the neural network model, and H (Z) represents entropy of the reconstruction coefficients. Wherein s.t. diag (Z) =0.

In the implementation, the neural network model is trained based on the training sample, model parameters in the neural network model are optimized through a gradient back propagation algorithm so that the model parameters meet an objective function, meanwhile, probability calculation is carried out on an output result of the neural network model through a softmax algorithm, the neural network model can be trained in the mode, and finally, the trained neural network model can be obtained.

In step S206, service timing information constructed by operation behavior information generated by the target user executing the target service for multiple times is obtained, where the service timing information includes information of the target service sequentially arranged according to a time sequence in which the target user executes the target service, and the attribute information of the target service sequentially arranged according to the time sequence includes operation behavior information generated by the target user executing the target service.

The specific processing of step S206 may be referred to the relevant content in the above embodiment, and will not be described herein.

In step S208, the service timing information is input into a pre-trained neural network model, and an output result corresponding to the service timing information is obtained.

In step S210, based on the output result corresponding to the service timing information, a reconstruction coefficient corresponding to the service timing information is determined by a softmax algorithm.

The softmax algorithm can ensure that the obtained reconstruction coefficient is a non-negative number and the sum of the reconstruction coefficients is 1, and the obtained reconstruction coefficient sum is a fixed value, so that the regularization term Z in the expression is zero ₁ Cannot be applied.

In implementation, the output result corresponding to the service time sequence information may be input into a formula corresponding to a softmax algorithm to obtain a corresponding calculation result, where the calculation result may be a reconstruction coefficient corresponding to the service time sequence information, i.e. z=softmax (f (X, Θ)), where f (X, Θ) represents a neural network model,

Representing traffic timing information.

In step S212, a matrix corresponding to the reconstruction coefficient corresponding to the service timing information and a transposed matrix of the matrix corresponding to the reconstruction coefficient are fused, so as to obtain a fusion matrix.

In practice, w=z+z may be used ^r Obtaining fusionA matrix W, wherein Z represents a matrix corresponding to the reconstruction coefficient corresponding to the service time sequence information, Z ^T Representing the transposed matrix of the matrix corresponding to the reconstruction coefficient, fusing the elements W in the matrix W _ij Describe x _i And x _j (j. Noteq. I) correlation, wherein a single or a small number of x can be used _j I.e. reconfigurable x _i 。

In step S214, an execution map of the target service is constructed based on the fusion matrix, where the execution map of the target service includes nodes and edges, the nodes are determined based on the target service executed each time in the service timing information, and the edges are determined based on the corresponding elements in the fusion matrix.

In implementation, as shown in fig. 3, the upper layer includes service timing information of the target service executed by the target user at 8 time points, and after the fusion matrix W is obtained in the above manner, the correlation W between two adjacent nodes (i.e. the target service executed at each time point) can be obtained _ij The fusion matrix can be used for constructing an execution map of the target service.

In step S216, the graph segmentation process is performed on the execution graph of the target service, so as to obtain the association relationship between the target services executed at different times in the service time sequence information.

In implementation, as shown in fig. 3, the middle layer and the lowest layer include service time sequence information formed by target services executed by a target user at 8 time points, and graph segmentation processing can be performed on an execution graph of the target services, where the middle layer performs graph segmentation once to obtain two sub-sequence information, and then performs graph segmentation on each sub-sequence information to obtain four sub-sequence information, so as to finally obtain an association relationship between the target services executed at different times in the service time sequence information.

For another example, as shown in fig. 4, the upper layer includes service time sequence information formed by the target service executed by the target user at 8 time points, after the fusion matrix W is obtained in the above manner, an execution spectrum of the target service may be constructed by using the fusion matrix, and a graph segmentation process may be performed on the execution spectrum of the target service to obtain two sub-sequence information, so as to finally obtain an association relationship between the target services executed at different times in the service time sequence information.

The processing in step S216 may be varied, and the following provides an alternative processing method, which may specifically include the following: and carrying out graph segmentation processing on the execution graph of the target service by a graph normalization segmentation algorithm to obtain the association relationship between the target services executed at different times in the service time sequence information.

The graph normalization segmentation algorithm can segment the graph based on the weight of the edges in the graph, cut off the edges with smaller weight (the weight is smaller than a preset threshold value), further form corresponding sub-graphs, and finally obtain the association relationship between the target businesses executed at different times in the business time sequence information.

In step S218, behavior state information of the target user for executing the target service is determined based on the association relationship between the target services executed at different times in the service timing information.

The specific process of step S218 may be referred to the relevant content in the above embodiment, and will not be described herein.

Example III

Based on the foregoing embodiment, the process of the foregoing embodiment is described below by using a specific application scenario, where the application scenario is a payment service scenario, and based on this, the target service is a payment service, the historical service timing information is historical payment timing information, and the service timing information is payment timing information.

As shown in fig. 5, the embodiment of the present disclosure provides a method for determining a user behavior state, where an execution subject of the method may be a terminal device or a server, where the terminal device may be a certain terminal device such as a mobile phone, a tablet computer, or a computer device such as a notebook computer or a desktop computer, or may also be an IoT device (specifically, such as a smart watch, an in-vehicle device, or the like). The server may be a single server, a server cluster including a plurality of servers, a background server such as a financial service or an online shopping service, or a background server of an application program. In this embodiment, a server is taken as an example for detailed description, and the following related contents may be referred to for the execution process of the terminal device, which is not described herein. The method specifically comprises the following steps:

In step S502, a training sample for training a neural network model is obtained, where the training sample is historical payment timing information constructed for historical operation behavior information generated by a user executing a payment service for multiple times, the historical payment timing information includes information of payment services sequentially arranged according to a time sequence in which the user executes the payment service, and attribute information of the payment services sequentially arranged according to the time sequence includes historical operation behavior information generated by the user executing the payment service.

The neural network model may include one or more of a recurrent neural network model, a convolutional neural network model, and an attention-mechanism-based neural network model, among others.

In step S504, model training is performed on the neural network model based on the training sample, and model parameters in the neural network model are optimized through a gradient back propagation algorithm, so that the model parameters satisfy an objective function, and a trained neural network model is obtained.

Wherein the objective function may be as follows:

the specific processing procedure can be referred to the above related content, and will not be described herein.

In step S506, payment timing information constructed by operation behavior information generated by the target user executing the payment service for a plurality of times is obtained, where the payment timing information includes information of the payment services sequentially arranged according to a time sequence in which the target user executes the payment service, and the attribute information of the payment services sequentially arranged according to the time sequence includes operation behavior information generated by the target user executing the payment service.

Wherein, the attribute information of the payment service sequentially arranged according to the time sequence comprises one or more of payment time, payment amount and position information.

In step S508, the payment timing information is input into a neural network model trained in advance, and an output result corresponding to the payment timing information is obtained.

In step S510, based on the output result corresponding to the above-mentioned payment timing information, the reconstruction coefficient corresponding to the payment timing information is determined by a softmax algorithm.

In step S512, a matrix corresponding to the reconstruction coefficient corresponding to the payment timing information and a transposed matrix of the matrix corresponding to the reconstruction coefficient are fused, so as to obtain a fusion matrix.

In step S514, an execution map of the payment service is constructed based on the fusion matrix, and the execution map of the payment service includes nodes and edges, the nodes are determined based on the payment service executed each time in the payment timing information, and the edges are determined based on the corresponding elements in the fusion matrix.

In step S516, the graph segmentation processing is performed on the execution graph of the payment service by the graph normalization segmentation algorithm, so as to obtain the association relationship between the payment services executed at different times in the payment time sequence information.

In step S518, behavior state information of the target user for executing the payment service is determined based on the association relationship between the payment services executed at different times in the above-mentioned payment timing information.

The determined behavior state information of the target user executing the payment service comprises one or more of whether the payment behavior of the target user is abnormal, whether the account of the target user is at risk or not, and different payment behavior states existing in the process of the target user executing the payment service.

Example IV

The above method for determining a user behavior state according to the embodiment of the present disclosure further provides a device for determining a user behavior state based on the same concept, as shown in fig. 6.

The device for determining the behavior state of the user comprises: a timing acquisition module 601, a reconstruction coefficient determination module 602, an association relationship determination module 603, and a behavior state determination module 604, wherein:

a time sequence acquisition module 601, configured to acquire service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, where the service time sequence information includes information of the target service sequentially arranged according to a time sequence in which the target user executes the target service, and attribute information of the target service sequentially arranged according to the time sequence includes operation behavior information generated by the target user executing the target service;

the reconstruction coefficient determining module 602 determines a reconstruction coefficient corresponding to the service time sequence information based on the service time sequence information and a pre-trained neural network model, and performs optimization processing on model parameters in the neural network model through the following objective functions in the process of training the neural network model: an objective function determined based on a norm constructed from training samples and reconstructed coefficient samples corresponding to the training samples, and entropy corresponding to the reconstructed coefficient samples;

The association relation determining module 603 determines association relation between target services executed at different times in the service timing information based on the reconstruction coefficient corresponding to the service timing information and the service timing information;

the behavior state determining module 604 determines behavior state information of the target user executing the target service based on an association relationship between the target services executed at different times in the service timing information.

In an embodiment of the present disclosure, the apparatus further includes:

the system comprises a sample acquisition module, a target service acquisition module and a target service acquisition module, wherein the sample acquisition module is used for acquiring training samples for training the neural network model, the training samples are historical service time sequence information constructed by historical operation behavior information generated by a user executing target services for a plurality of times, the historical service time sequence information comprises information of the target services which are sequentially arranged according to the time sequence of the user executing the target services, and attribute information of the target services which are sequentially arranged according to the time sequence comprises historical operation behavior information generated by the user executing the target services;

and the model training module is used for carrying out model training on the neural network model based on the training sample, and carrying out optimization processing on model parameters in the neural network model through a gradient back propagation algorithm so that the model parameters meet the objective function, thereby obtaining the trained neural network model.

In the embodiment of the present disclosure, the reconstruction coefficient determining module 602 includes:

the model processing unit inputs the service time sequence information into a pre-trained neural network model to obtain an output result corresponding to the service time sequence information;

and the reconstruction coefficient determining unit is used for determining the reconstruction coefficient corresponding to the service time sequence information through a softmax algorithm based on the output result corresponding to the service time sequence information.

In the embodiment of the present disclosure, the association determining module 603 includes:

the fusion unit is used for carrying out fusion processing on a matrix corresponding to the reconstruction coefficient corresponding to the service time sequence information and a transposed matrix of the matrix corresponding to the reconstruction coefficient to obtain a fusion matrix;

the map construction unit is used for constructing an execution map of the target service based on the fusion matrix, wherein the execution map of the target service comprises nodes and edges, the nodes are determined based on the target service executed each time in the service time sequence information, and the edges are determined based on corresponding elements in the fusion matrix;

and the graph segmentation unit is used for performing graph segmentation processing on the execution graph of the target service to obtain the association relationship between the target services executed at different times in the service time sequence information.

In this embodiment of the present disclosure, the graph segmentation unit performs graph segmentation processing on the execution graph of the target service by using a graph normalization segmentation algorithm, so as to obtain an association relationship between target services executed at different times in the service timing information.

In embodiments of the present disclosure, the neural network model includes one or more of a recurrent neural network model, a convolutional neural network model, and an attention-based neural network model.

In this embodiment of the present disclosure, the target service is a payment service, and the attribute information of the target service sequentially arranged according to a time sequence includes one or more of payment time, payment amount, and location information; the determined behavior state information of the target user executing the target service comprises one or more of whether the payment behavior of the target user is abnormal, whether the account of the target user is at risk or not, and different payment behavior states existing in the process of executing the payment service by the target user.

The embodiment of the specification provides a determining device for a user behavior state, which is configured by acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, attribute information of the target service which is sequentially arranged according to the time sequence comprises operation behavior information generated by the target user executing the target service, then, based on the service time sequence information and a pre-trained neural network model, determining a reconstruction coefficient corresponding to the service time sequence information, and optimizing model parameters in the neural network model through the following target functions in the process of training the neural network model: based on the norm constructed by the training samples and the reconstructed coefficient samples corresponding to the training samples and the target function determined by the entropy corresponding to the reconstructed coefficient samples, then, the association relationship between the target services executed at different times in the service time sequence information can be determined based on the reconstructed coefficient corresponding to the service time sequence information and the service time sequence information, finally, the behavior state information of the target user executing the target service is determined based on the association relationship between the target services executed at different times in the service time sequence information, thus, the reconstructed coefficient corresponding to the service time sequence information is determined based on the pre-trained neural network model, the correlation relationship of the data in the service time sequence information is further extracted and clustering processing is carried out, the neural network model can output probability weight (namely, the reconstructed coefficient corresponding to the service time sequence information) as an encoder, the calculation process of the neural network model is faster than that of the optimization algorithm based on the sparse coding extracted data correlation relationship, and the hardware acceleration processing is easy to use GPU and the like, so that different behavior states of different users can be determined more quickly and effectively under the condition of lack of calibration, and the behavior state of different users can be determined.

Example five

The above determining device for a user behavior state provided in the embodiment of the present disclosure further provides a determining device for a user behavior state based on the same concept, as shown in fig. 7.

The user behavior state determining device may provide a terminal device or a server or the like for the above-described embodiments.

The device for determining the behavior state of the user may have a relatively large difference due to different configurations or performances, and may include one or more processors 701 and a memory 702, where the memory 702 may store one or more stored applications or data. Wherein the memory 702 may be transient storage or persistent storage. The application programs stored in the memory 702 may include one or more modules (not shown in the figures), each of which may include a series of computer-executable instructions in the device for determining the behavior state of a user. Still further, the processor 701 may be configured to communicate with the memory 702 and execute a series of computer executable instructions in the memory 702 on the device for determining the user behavior state. The device for determining the status of a user's behavior may also include one or more power supplies 703, one or more wired or wireless network interfaces 704, one or more input/output interfaces 705, one or more keyboards 706.

In particular, in this embodiment, the determining device of the user behavior state includes a memory, and one or more programs, where the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer executable instructions in the determining device of the user behavior state, and executing the one or more programs by the one or more processors includes computer executable instructions for:

acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, and the attribute information of the target service which is sequentially arranged according to the time sequence comprises the operation behavior information generated by the target user executing the target service;

based on the service time sequence information and a pre-trained neural network model, determining a reconstruction coefficient corresponding to the service time sequence information, and optimizing model parameters in the neural network model in the process of training the neural network model through the following objective functions: an objective function determined based on a norm constructed from training samples and reconstructed coefficient samples corresponding to the training samples, and entropy corresponding to the reconstructed coefficient samples;

Determining the association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information;

and determining behavior state information of the target user executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

In this embodiment of the present specification, further includes:

acquiring a training sample for training the neural network model, wherein the training sample is historical service time sequence information constructed by historical operation behavior information generated by a user executing target services for a plurality of times, the historical service time sequence information comprises information of the target services which are sequentially arranged according to the time sequence of the user executing the target services, and the attribute information of the target services which are sequentially arranged according to the time sequence comprises historical operation behavior information generated by the user executing the target services;

model training is carried out on the neural network model based on the training sample, and model parameters in the neural network model are optimized through a gradient back propagation algorithm, so that the model parameters meet the objective function, and a trained neural network model is obtained.

In this embodiment of the present disclosure, the determining, based on the service timing information and a neural network model trained in advance, a reconstruction coefficient corresponding to the service timing information includes:

inputting the service time sequence information into a pre-trained neural network model to obtain an output result corresponding to the service time sequence information;

and determining a reconstruction coefficient corresponding to the service time sequence information through a softmax algorithm based on an output result corresponding to the service time sequence information.

In this embodiment of the present disclosure, the determining, based on the reconstruction coefficient corresponding to the service timing information and the service timing information, an association relationship between target services executed at different times in the service timing information includes:

fusing a matrix corresponding to the reconstruction coefficient corresponding to the service time sequence information with a transposed matrix of the matrix corresponding to the reconstruction coefficient to obtain a fusion matrix;

constructing an execution map of the target service based on the fusion matrix, wherein the execution map of the target service comprises nodes and edges, the nodes are determined based on the target service executed each time in the service time sequence information, and the edges are determined based on corresponding elements in the fusion matrix;

And carrying out graph segmentation processing on the execution graph of the target service to obtain the association relation between the target services executed at different times in the service time sequence information.

In this embodiment of the present disclosure, performing graph segmentation on the execution graph of the target service to obtain an association relationship between target services executed at different times in the service timing information includes:

and carrying out graph segmentation processing on the execution graph of the target service by a graph normalization segmentation algorithm to obtain the association relationship between the target services executed at different times in the service time sequence information.

The embodiment of the specification provides a determining device for a user behavior state, which is configured by acquiring service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, attribute information of the target service which is sequentially arranged according to the time sequence comprises operation behavior information generated by the target user executing the target service, then, based on the service time sequence information and a pre-trained neural network model, a reconstruction coefficient corresponding to the service time sequence information is determined, and model parameters in the neural network model are optimized by the following objective functions in the process of training the neural network model: based on the norm constructed by the training samples and the reconstructed coefficient samples corresponding to the training samples and the target function determined by the entropy corresponding to the reconstructed coefficient samples, then, the association relationship between the target services executed at different times in the service time sequence information can be determined based on the reconstructed coefficient corresponding to the service time sequence information and the service time sequence information, finally, the behavior state information of the target user executing the target service is determined based on the association relationship between the target services executed at different times in the service time sequence information, thus, the reconstructed coefficient corresponding to the service time sequence information is determined based on the pre-trained neural network model, the correlation relationship of the data in the service time sequence information is further extracted and clustering processing is carried out, the neural network model can output probability weight (namely, the reconstructed coefficient corresponding to the service time sequence information) as an encoder, the calculation process of the neural network model is faster than that of the optimization algorithm based on the sparse coding extracted data correlation relationship, and the hardware acceleration processing is easy to use GPU and the like, so that different behavior states of different users can be determined more quickly and effectively under the condition of lack of calibration, and the behavior state of different users can be determined.

Example six

Further, based on the method shown in fig. 1 to 5, one or more embodiments of the present disclosure further provide a storage medium, which is used to store computer executable instruction information, and in a specific embodiment, the storage medium may be a U disc, an optical disc, a hard disk, etc., where the computer executable instruction information stored in the storage medium can implement the following flow when executed by a processor:

In this embodiment of the present specification, further includes:

The embodiment of the present disclosure provides a storage medium, configured to obtain service timing information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, where the service timing information includes information of target services sequentially arranged according to a time sequence in which the target user executes the target service, and attribute information of the target services sequentially arranged according to the time sequence includes operation behavior information generated by the target user executing the target service, and then determine a reconstruction coefficient corresponding to the service timing information based on the service timing information and a pre-trained neural network model, and perform optimization processing on model parameters in the neural network model by using the following objective functions in a process of training the neural network model: based on the norm constructed by the training samples and the reconstructed coefficient samples corresponding to the training samples and the target function determined by the entropy corresponding to the reconstructed coefficient samples, then, the association relationship between the target services executed at different times in the service time sequence information can be determined based on the reconstructed coefficient corresponding to the service time sequence information and the service time sequence information, finally, the behavior state information of the target user executing the target service is determined based on the association relationship between the target services executed at different times in the service time sequence information, thus, the reconstructed coefficient corresponding to the service time sequence information is determined based on the pre-trained neural network model, the correlation relationship of the data in the service time sequence information is further extracted and clustering processing is carried out, the neural network model can output probability weight (namely, the reconstructed coefficient corresponding to the service time sequence information) as an encoder, the calculation process of the neural network model is faster than that of the optimization algorithm based on the sparse coding extracted data correlation relationship, and the hardware acceleration processing is easy to use GPU and the like, so that different behavior states of different users can be determined more quickly and effectively under the condition of lack of calibration, and the behavior state of different users can be determined.

The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable Gate Array, FPGA)) is an integrated circuit whose logic function is determined by the programming of the device by a user. A designer programs to "integrate" a digital system onto a PLD without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented by using "logic compiler" software, which is similar to the software compiler used in program development and writing, and the original code before the compiling is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but not just one of the hdds, but a plurality of kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), lava, lola, myHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers, and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller may thus be regarded as a kind of hardware component, and means for performing various functions included therein may also be regarded as structures within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.

The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functionality of the units may be implemented in one or more software and/or hardware when implementing one or more embodiments of the present description.

It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

Embodiments of the present description are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable fraud case serial-to-parallel device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable fraud case serial-to-parallel device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.

One or more embodiments of the present specification may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the present description may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.

The foregoing is merely exemplary of the present disclosure and is not intended to limit the present disclosure. Various modifications and alterations to this specification will become apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, or the like, which are within the spirit and principles of the present description, are intended to be included within the scope of the claims of the present description.

Claims

1. A method of determining a user behavior state, the method comprising:

2. The method of claim 1, the method further comprising:

3. The method of claim 1, wherein the determining, based on the traffic timing information and a pre-trained neural network model, a reconstruction coefficient corresponding to the traffic timing information comprises:

4. The method of claim 1, wherein the determining the association relationship between the target services executed at different times in the service timing information based on the reconstruction coefficient corresponding to the service timing information and the service timing information comprises:

5. The method of claim 4, wherein performing graph segmentation on the execution graph of the target service to obtain the association relationship between the target services executed at different times in the service timing information, comprises:

6. The method of any of claims 1-5, the neural network model comprising one or more of a recurrent neural network model, a convolutional neural network model, an attention-mechanism-based neural network model.

7. The method of claim 6, wherein the target service is a payment service, and the attribute information of the target service sequentially arranged in time sequence includes one or more of payment time, payment amount, and location information; the determined behavior state information of the target user executing the target service comprises one or more of whether the payment behavior of the target user is abnormal, whether the account of the target user is at risk or not, and different payment behavior states existing in the process of executing the payment service by the target user.

8. A device for determining a user behavior state, the device comprising:

the time sequence acquisition module acquires service time sequence information constructed by operation behavior information generated by a target user executing a target service for a plurality of times, wherein the service time sequence information comprises information of the target service which is sequentially arranged according to the time sequence of the target user executing the target service, and the attribute information of the target service which is sequentially arranged according to the time sequence comprises the operation behavior information generated by the target user executing the target service;

the reconstruction coefficient determining module is used for determining a reconstruction coefficient corresponding to the service time sequence information based on the service time sequence information and a pre-trained neural network model, and optimizing model parameters in the neural network model through the following objective functions in the process of training the neural network model: an objective function determined based on a norm constructed from training samples and reconstructed coefficient samples corresponding to the training samples, and entropy corresponding to the reconstructed coefficient samples;

the association relation determining module is used for determining association relation between target businesses executed at different times in the business time sequence information based on the reconstruction coefficient corresponding to the business time sequence information and the business time sequence information;

And the behavior state determining module is used for determining the behavior state information of the target user executing the target service based on the association relation between the target services executed at different times in the service time sequence information.

9. A user behavior state determination device, the user behavior state determination device comprising:

a processor; and

a memory arranged to store computer executable instructions that, when executed, cause the processor to:

10. A storage medium for storing computer executable instructions that when executed by a processor implement the following: