A construction worker management method and system

Abstract

The present invention relates to the technical field of personnel management, and in particular to a construction worker management method and system. A construction worker management system comprises: a construction worker behavior data acquisition module and an access behavior threat analysis module. The present invention analyzes the historical access data of construction workers through an access behavior threat analysis model, and determines whether the current access behavior poses a threat of data leakage, and timely interrupts the access behavior of construction workers in the event of a data leakage threat, to avoid further data leakage, thereby achieving access behavior management for construction workers; in the access behavior threat analysis model, a dependency weight matrix is ​​first obtained through dependency analysis, and then the dependency weight matrix is ​​used to provide a reference for subsequent behavior data prediction, so as to fully learn the changing trend of historical access data and the dependency relationship between data items, and improve the accuracy of access behavior threat analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of personnel management, and in particular to a construction worker management method and system. Background Art

[0002] With the digital transformation of the construction industry, more and more construction projects have begun to introduce information management systems to improve project management efficiency, optimize resource scheduling, and improve overall construction quality. In these systems, construction workers access the internal systems of construction projects to obtain relevant construction plans, design drawings, bills of materials and other key data to ensure the smooth progress of the construction process. However, with the increasing degree of informatization, construction workers face increasingly severe data security issues when accessing these systems. Summary of the invention

[0003] The present invention analyzes the historical access data of construction workers through an access behavior threat analysis model, determines whether the current access behavior poses a threat of data leakage, and promptly interrupts the access behavior of construction workers when there is a threat of data leakage to avoid further data leakage, thereby achieving access behavior management for construction workers; in the access behavior threat analysis model, a dependency weight matrix is ​​first obtained through dependency analysis, and then the dependency weight matrix is ​​used to provide a reference for subsequent behavior data prediction, which can fully learn the changing trends of historical access data and the dependencies between data items, and improve the accuracy of access behavior threat analysis.

[0004] The present invention provides a construction worker management method, comprising:

[0005] Whenever a construction worker performs an access behavior, obtaining construction worker behavior data of the construction worker;

[0006] The N construction worker behavior data before the current monitoring time point are combined into a construction worker behavior analysis data set, where N is the total number of construction worker behavior data in the construction worker behavior analysis data set. The construction worker behavior analysis data set and the current construction worker behavior data are then sent to the access behavior threat analysis model for processing, and an access behavior threat label is output. The access behavior threat label includes threat and no threat. If the output access behavior threat label is threat, all access behaviors of the construction worker are interrupted to avoid further data leakage, and a dynamic password for executing access behavior is randomly generated for the construction worker; if the output access behavior threat label is not threat, no operation is performed to achieve management of the construction worker;

[0007] The access behavior threat analysis model includes a dependency weight matrix construction layer, a behavior data prediction layer, a feature enhancement layer, a behavior data analysis layer and an access behavior threat label output layer; wherein the dependency weight matrix construction layer is used to perform dependency analysis on the construction worker behavior analysis data set and construct a dependency weight matrix; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set based on the dependency weight matrix and construct predicted construction worker behavior data; the feature enhancement layer is used to perform feature enhancement on the construction worker behavior data acquired at the current monitoring time point based on the predicted construction worker behavior data and construct the construction worker behavior data to be analyzed; the behavior data analysis layer is used to analyze the construction worker behavior data to be analyzed and generate access behavior threat labels; the access behavior threat label output layer is used to output access behavior threat labels;

[0008] Preferably, the dependency analysis of the construction worker behavior analysis data set is performed through the dependency weight matrix construction layer to construct a dependency weight matrix, which specifically includes the following steps:

[0009] Construct N-1 sets of construction worker behavior analysis differential data. The construction method of the construction worker behavior analysis differential data is as follows: perform a differential operation on the n+1th construction worker behavior data in the construction worker behavior analysis data set and the nth construction worker behavior data in the construction worker behavior analysis data set to construct the nth construction worker behavior analysis differential data, n=1, 2, 3, ..., N-1;

[0010] Traverse N-1 construction worker behavior analysis differential data, and for the selected i-th construction worker behavior analysis differential data, i=1, 2, 3, ..., N-2, perform the following operations: multiply the i-th construction worker behavior analysis differential data with the differential key matrix to construct the construction worker behavior analysis key matrix K, multiply the i+1-th construction worker behavior analysis differential data with the differential query key matrix to construct the construction worker behavior analysis query matrix Q, and calculate the differential attention weight matrix A i =softmax(QK T ), where T is the transposition operation; until N-1 construction workers' behavior analysis differential data are all traversed, the N-1 differential attention weight matrix A i Perform an averaging operation to construct a dependency weight matrix.

[0011] Preferably, in the behavior data prediction layer, the behavior data of the construction worker behavior analysis data set is predicted based on the dependency weight matrix to construct the predicted construction worker behavior data, which specifically includes the following steps:

[0012] The behavior data prediction layer is built based on the LSTM model. The difference from the traditional LSTM model is that every time a hidden vector is generated in the behavior data prediction layer, the hidden vector is updated with the dependency weight matrix. The behavior data prediction layer is used to predict the behavior data of the construction workers' behavior analysis dataset and construct the predicted construction workers' behavior data.

[0013] The hidden vector is updated with the dependency weight matrix, which specifically includes the following steps: multiplying the hidden vector with the hidden key matrix and the hidden value matrix respectively to construct the behavior data hidden key matrix H and the behavior data hidden value matrix G; multiplying the dependency weight matrix with the hidden query key matrix to construct the behavior data hidden query matrix R; and calculating the hidden attention weight matrix B = softmax (RH T ), and then multiply the behavior data hidden value matrix G with the hidden attention weight matrix B to update the hidden vector.

[0014] Preferably, the construction worker behavior data acquired at the current monitoring time point is feature enhanced through the feature enhancement layer to construct the construction worker behavior data to be analyzed, which specifically includes the following steps:

[0015] Performing a differential operation on the construction worker behavior data obtained at the current monitoring time point and the predicted construction worker behavior data to construct the construction worker behavior differential data, and enhancing the features of the construction worker behavior differential data through the dependency weight matrix to construct the construction worker behavior data to be analyzed;

[0016] The feature enhancement of construction workers' behavior differential data is performed by relying on the weight matrix, which specifically includes the following steps: multiplying the construction workers' behavior differential data with the enhancement key matrix and the enhancement value matrix respectively, constructing the behavior data enhancement key matrix X and the behavior data enhancement value matrix Y, multiplying the dependency weight matrix with the enhancement query key matrix, constructing the behavior data enhancement query matrix E, and calculating the enhancement attention weight matrix C = softmax (EX T ), and then multiply the behavior data reinforcement value matrix Y and the reinforcement attention weight matrix C to enhance the features of the construction workers' behavior differential data and construct the construction workers' behavior data to be analyzed.

[0017] Preferably, training is performed on the access behavior threat analysis model, specifically including the following steps:

[0018] Obtain several positive training samples and negative training samples for access behavior threat analysis, wherein the positive training samples for access behavior threat analysis include N+1 pieces of construction worker behavior data arranged in chronological order without data leakage, and the negative training samples for access behavior threat analysis include N+1 pieces of construction worker behavior data arranged in chronological order with data leakage, and the last piece of construction worker behavior data in the negative training samples for access behavior threat analysis is obtained when data leakage occurs, annotate the positive training samples for access behavior threat analysis with non-threatening access behavior threat labels, annotate the negative training samples for access behavior threat analysis with threatening access behavior threat labels, and All labeled access behavior threat analysis positive training samples and access behavior threat analysis negative training samples constitute the access behavior threat analysis training set. The access behavior threat analysis model with parameter initialization is trained through the access behavior threat analysis training set, and the loss value is calculated. Based on the loss value, the parameters in the access behavior threat analysis model are adjusted through the back propagation algorithm to obtain the accuracy of the access behavior threat analysis model, and determine whether the accuracy of the access behavior threat analysis model is higher than the preset threshold. If the accuracy of the access behavior threat analysis model is higher than the preset threshold, the trained access behavior threat analysis model is output; otherwise, the access behavior threat analysis model continues to be trained through the access behavior threat analysis training set.

[0019] Preferably, the behavior data analysis layer is established based on BP neural network.

[0020] Preferably, the hyperparameters in the access behavior threat analysis model are simulated and set by a swarm optimization algorithm.

[0021] The present invention also provides a construction worker management system, comprising:

[0022] A construction worker behavior data acquisition module, used for acquiring construction worker behavior data of the construction worker whenever the construction worker performs an access behavior;

[0023] The access behavior threat analysis module is used to form a construction worker behavior analysis data set with N pieces of construction worker behavior data before the current monitoring time point, where N is the total number of construction worker behavior data in the construction worker behavior analysis data set, and then send the construction worker behavior analysis data set and the current construction worker behavior data to the access behavior threat analysis model for processing, and output an access behavior threat label, which includes threat and no threat. If the output access behavior threat label is threat, all access behaviors of the construction worker are interrupted to avoid further leakage of data, and a dynamic password for executing access behavior is randomly generated for the construction worker; if the output access behavior threat label is not threat, no operation is performed to achieve management of the construction worker;

[0024] The access behavior threat analysis model includes a dependency weight matrix construction layer, a behavior data prediction layer, a feature enhancement layer, a behavior data analysis layer and an access behavior threat label output layer; wherein the dependency weight matrix construction layer is used to perform dependency analysis on the construction worker behavior analysis data set and construct a dependency weight matrix; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set based on the dependency weight matrix and construct predicted construction worker behavior data; the feature enhancement layer is used to perform feature enhancement on the construction worker behavior data acquired at the current monitoring time point based on the predicted construction worker behavior data and construct the construction worker behavior data to be analyzed; the behavior data analysis layer is used to analyze the construction worker behavior data to be analyzed and generate access behavior threat labels; the access behavior threat label output layer is used to output access behavior threat labels.

[0025] The present invention has the following advantages:

[0026] The present invention analyzes the historical access data of construction workers through an access behavior threat analysis model, determines whether the current access behavior poses a threat of data leakage, and promptly interrupts the access behavior of construction workers when there is a threat of data leakage to avoid further data leakage, thereby achieving access behavior management for construction workers; in the access behavior threat analysis model, a dependency weight matrix is ​​first obtained through dependency analysis, and then the dependency weight matrix is ​​used to provide a reference for subsequent behavior data prediction, which can fully learn the changing trends of historical access data and the dependencies between data items, and improve the accuracy of access behavior threat analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic diagram of the structure of a construction worker management system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0029] Embodiment 1, a construction worker management method, comprising:

[0030] Whenever a construction worker performs an access behavior, the construction worker behavior data of the construction worker is obtained. The construction worker behavior data includes the login behavior (such as login time, login frequency, login device, login IP address and geographic location, etc.) of the construction worker when accessing the internal system of the construction project, access behavior (accessed file type, access frequency, access duration and sensitive data accessed, etc.), network status (such as network type, network security and connected IP address, etc.), operation habits (such as mouse movement trajectory, keyboard input speed and touch screen operation frequency, etc.), these construction worker behavior data can reflect the behavioral characteristics of the construction worker when accessing the construction project; and these construction worker behavior data can be obtained through log files and network monitoring and other methods; it should be noted that in the process of executing a construction project, a construction project internal system is generally used to manage the data generated in the construction project process, such as construction worker attendance data, construction project progress data and construction worker personal file data, etc.;

[0031] The N construction worker behavior data before the current monitoring time point are combined into a construction worker behavior analysis data set, where N is the total number of construction worker behavior data in the construction worker behavior analysis data set. The construction worker behavior analysis data set and the current construction worker behavior data are then sent to the access behavior threat analysis model for processing, and an access behavior threat label is output. The access behavior threat label includes threat and no threat. If the output access behavior threat label is threat, it means that the currently acquired construction worker behavior data does not conform to the construction worker's historical access behavior, and there is a risk of data leakage. For example, if a stranger pretends to be a construction worker to access the internal system of a construction project, all access behaviors of the construction worker are interrupted to avoid further data leakage, and a dynamic password for executing access behavior is randomly generated for the construction worker, and the real construction worker can obtain the generated dynamic password through biometric recognition and other methods; if the output access behavior threat label is not threat, no operation is performed to achieve management of construction workers;

[0032] The access behavior threat analysis model includes a dependency weight matrix construction layer, a behavior data prediction layer, a feature enhancement layer, a behavior data analysis layer and an access behavior threat label output layer; wherein the dependency weight matrix construction layer is used to perform dependency analysis on the construction worker behavior analysis data set and construct a dependency weight matrix. It should be noted that in the process of performing dependency analysis on the construction worker behavior analysis data set, the mutual dependency of each item of data in the construction worker behavior data can be analyzed, such as the correspondence between the daytime working period and the construction address, to provide prior knowledge for subsequent behavior prediction; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set based on the dependency weight matrix and construct predicted construction worker behavior data. It should be noted that with the assistance of the dependency weight matrix, it is possible to combine the prior behavior data item dependencies and improve the accuracy of behavior data prediction; the feature enhancement layer is used to predict the construction worker behavior data acquired at the current monitoring time point based on the predicted construction worker behavior data. Feature enhancement is performed to construct the construction worker behavior data to be analyzed. In the process of feature enhancement, with the assistance of predicting the construction worker behavior data, the abnormal data items of the construction worker behavior data obtained at the current monitoring time point can be enhanced to improve the accuracy of subsequent behavior data analysis; the behavior data analysis layer is established based on the BP neural network, which specifically includes an input layer, three hidden layers and an output layer, and the input layer, the hidden layer and the output layer are all constructed by a number of neural nodes. The number of neural nodes in the input layer and the output layer is consistent with the dimension of the construction worker behavior data to be analyzed. The neural nodes of the hidden layer are set by the developer. The behavior data analysis layer is used to analyze the construction worker behavior data to be analyzed and generate access behavior threat labels, wherein the input layer is used to receive the construction worker behavior data to be analyzed, the hidden layer is used to analyze the construction worker behavior data to be analyzed, and the output layer is used to output the access behavior threat labels; the access behavior threat label output layer is used to output the access behavior threat labels;

[0033] The present application analyzes the historical access data of construction workers through an access behavior threat analysis model, determines whether the current access behavior poses a threat of data leakage, and promptly interrupts the access behavior of construction workers when there is a threat of data leakage to avoid further data leakage, thereby achieving management of the access behavior of construction workers; in the access behavior threat analysis model, a dependency weight matrix is ​​first obtained through dependency analysis, and then the dependency weight matrix is ​​used to provide a reference for subsequent behavior data prediction, which can fully learn the changing trends of historical access data and the dependencies between data items, and improve the accuracy of access behavior threat analysis.

[0034] The dependency analysis of the construction worker behavior analysis dataset is performed through the dependency weight matrix construction layer to construct a dependency weight matrix, which specifically includes the following steps:

[0035] Construct N-1 sets of construction worker behavior analysis differential data. The construction method of construction worker behavior analysis differential data is: perform a differential operation on the n+1th construction worker behavior data in the construction worker behavior analysis data set and the nth construction worker behavior data in the construction worker behavior analysis data set to construct the nth construction worker behavior analysis differential data, n=1, 2, 3, ..., N-1. The construction worker behavior analysis differential data can reflect the changes in data items between two adjacent construction worker behavior data on the timeline. Data items with consistent change trends can be regarded as having a dependency relationship;

[0036] Traverse N-1 construction worker behavior analysis differential data, and for the selected i-th construction worker behavior analysis differential data, i=1, 2, 3, ..., N-2, perform the following operations: multiply the i-th construction worker behavior analysis differential data with the differential key matrix to construct the construction worker behavior analysis key matrix K, multiply the i+1-th construction worker behavior analysis differential data with the differential query key matrix to construct the construction worker behavior analysis query matrix Q, and calculate the differential attention weight matrix A i =softmax(QK T ), where T is the transposition operation. The self-attention mechanism performed by the differential key matrix and the query key matrix can analyze the dependency of the changes between the data items, and the differential key matrix and the differential query key matrix are set with reference to the self-attention mechanism in the Transformer model, and the parameters of the differential key matrix and the differential query key matrix are adjusted with the training of the access behavior threat analysis model; until the N-1 construction workers' behavior analysis differential data are traversed, the N-1 differential attention weight matrices A i Perform an averaging operation to construct a dependency weight matrix.

[0037] In the behavior data prediction layer, the behavior data of the construction worker behavior analysis dataset is predicted based on the dependency weight matrix to construct the predicted construction worker behavior data, which specifically includes the following steps:

[0038] The behavior data prediction layer is established based on the LSTM model. The traditional LSTM model includes an input gate, a forget gate, a memory unit and an output gate. The specific setting of the behavior data prediction layer refers to the LSTM model in the prior art and will not be described here. The difference from the traditional LSTM model is that whenever a hidden vector is generated in the behavior data prediction layer, the hidden vector is updated with a dependency weight matrix. It should be noted that the hidden vector generated in the behavior data prediction layer is specifically a hidden vector stored after the memory unit is updated by the forget gate and the hidden gate in the behavior data prediction layer; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set to construct predicted construction worker behavior data;

[0039] The hidden vector is updated with the dependency weight matrix, which specifically includes the following steps: multiplying the hidden vector with the hidden key matrix and the hidden value matrix respectively to construct the behavior data hidden key matrix H and the behavior data hidden value matrix G; multiplying the dependency weight matrix with the hidden query key matrix to construct the behavior data hidden query matrix R; and calculating the hidden attention weight matrix B = softmax (RH T ), the dependency weight matrix can be used to enhance the features in the hidden vector with the help of the data item dependency relationship, and then the behavior data hidden value matrix G is multiplied by the hidden attention weight matrix B to update the hidden vector; and the hidden key matrix, hidden value matrix and hidden query key matrix are all set with reference to the self-attention mechanism in the Transformer model, and the parameters of the hidden key matrix, hidden value matrix and hidden query key matrix are adjusted with the training of the access behavior threat analysis model;

[0040] The feature enhancement layer is used to enhance the features of the construction worker behavior data obtained at the current monitoring time point, and the construction worker behavior data to be analyzed is constructed, which specifically includes the following steps:

[0041] Performing a differential operation on the construction worker behavior data obtained at the current monitoring time point and the predicted construction worker behavior data to construct the construction worker behavior differential data, and enhancing the features of the construction worker behavior differential data through the dependency weight matrix to construct the construction worker behavior data to be analyzed;

[0042] The feature enhancement of construction workers' behavior differential data is performed by relying on the weight matrix, which specifically includes the following steps: multiplying the construction workers' behavior differential data with the enhancement key matrix and the enhancement value matrix respectively, constructing the behavior data enhancement key matrix X and the behavior data enhancement value matrix Y, multiplying the dependency weight matrix with the enhancement query key matrix, constructing the behavior data enhancement query matrix E, and calculating the enhancement attention weight matrix C = softmax (EX T ), with the help of the dependency weight matrix, the features in the construction workers' behavior differential data can be strengthened with the help of the data item dependency relationship, and then the behavior data reinforcement value matrix Y and the reinforcement attention weight matrix C are multiplied to strengthen the features of the construction workers' behavior differential data and construct the construction workers' behavior data to be analyzed; and the reinforcement key matrix, reinforcement value matrix and reinforcement query key matrix are all set with reference to the self-attention mechanism in the Transformer model, and the parameters of the reinforcement key matrix, reinforcement value matrix and reinforcement query key matrix are adjusted with the training of the access behavior threat analysis model;

[0043] Training the access behavior threat analysis model includes the following steps:

[0044] Obtain several positive training samples and negative training samples for access behavior threat analysis. The positive training samples for access behavior threat analysis include N+1 pieces of construction worker behavior data arranged in chronological order without data leakage, and the negative training samples for access behavior threat analysis include N+1 pieces of construction worker behavior data arranged in chronological order with data leakage, and the last piece of construction worker behavior data in the negative training samples for access behavior threat analysis is obtained when data leakage occurs. It should be noted that the construction worker behavior data in the positive training samples and the negative training samples for access behavior threat analysis are all obtained through actual data access experiments. The positive training samples for access behavior threat analysis are annotated with non-threatening access behavior threat labels, and the negative training samples for access behavior threat analysis are annotated with threatening access behavior threat labels. Labeling, and forming an access behavior threat analysis training set with all labeled access behavior threat analysis positive training samples and access behavior threat analysis negative training samples, training the access behavior threat analysis model with initialized parameters through the access behavior threat analysis training set, calculating the loss value through the cross entropy loss function, adjusting the parameters in the access behavior threat analysis model through the back propagation algorithm based on the loss value, obtaining the accuracy of the access behavior threat analysis model, and judging whether the accuracy of the access behavior threat analysis model is higher than the preset threshold value, which is set by the developer. If the accuracy of the access behavior threat analysis model is higher than the preset threshold value, output the trained access behavior threat analysis model; otherwise, continue to train the access behavior threat analysis model through the access behavior threat analysis training set; and the hyperparameters in the access behavior threat analysis model are simulated and set through the swarm optimization algorithm;

[0045] The hyperparameters in the access behavior threat analysis model are simulated and set through a swarm optimization algorithm, which specifically includes the following steps:

[0046] Construct several simulated individuals, which store the hyperparameters in the access behavior threat analysis model. During the construction of the simulated individuals, a value within the corresponding range is randomly selected for each hyperparameter, and all simulated individuals are grouped into a population set, and the maximum number of iterations is set;

[0047] Calculate the fitness of each simulated individual. The specific calculation method is as follows: set the hyperparameters in the access behavior threat analysis model through the simulated individual, and train the access behavior threat analysis model through the access behavior threat analysis training set, and use the accuracy of the access behavior threat analysis model as the fitness of the simulated individual.

[0048] Based on the fitness of the simulated individuals, the population set is iteratively updated through the sparrow search algorithm;

[0049] Until the number of iterations reaches the maximum number of iterations, the simulated individual with the largest fitness is output to set the hyperparameters in the access behavior threat analysis model.

[0050] Embodiment 2, a construction worker management system, such as Figure 1 As shown, including:

[0051] The construction worker behavior data acquisition module is used to acquire the construction worker behavior data of the construction worker when the construction worker performs access behavior. The construction worker behavior data includes the login behavior (such as login time, login frequency, login device, login IP address and geographical location, etc.) when the construction worker accesses the internal system of the construction project, access behavior (accessed file type, access frequency, access duration and sensitive data accessed, etc.), network status (such as network type, network security and connected IP address, etc.), operation habits (such as mouse movement trajectory, keyboard input speed and touch screen operation frequency, etc.), these construction worker behavior data can reflect the behavioral characteristics of the construction worker when accessing the construction project; and these construction worker behavior data can be obtained through log files and network monitoring and other methods; it should be noted that in the process of executing the construction project, a construction project internal system is generally used to manage the data generated in the construction project process, such as construction worker attendance data, construction project progress data and construction worker personal file data;

[0052] The access behavior threat analysis module is used to form a construction worker behavior analysis data set with N construction worker behavior data before the current monitoring time point, where N is the total number of construction worker behavior data in the construction worker behavior analysis data set, and then send the construction worker behavior analysis data set and the current construction worker behavior data to the access behavior threat analysis model for processing, and output an access behavior threat label. The access behavior threat label includes threat and no threat. If the output access behavior threat label is threat, it means that the currently acquired construction worker behavior data does not conform to the construction worker's historical access behavior, and there is a risk of data leakage. For example, if a stranger pretends to be a construction worker to access the internal system of a construction project, all access behaviors of the construction worker are interrupted to avoid further data leakage, and a dynamic password for executing access behavior is randomly generated for the construction worker, and the real construction worker can obtain the generated dynamic password through biometric recognition and other methods; if the output access behavior threat label is not threat, no operation is performed to achieve management of the construction worker;

[0053] The access behavior threat analysis model includes a dependency weight matrix construction layer, a behavior data prediction layer, a feature enhancement layer, a behavior data analysis layer and an access behavior threat label output layer; wherein the dependency weight matrix construction layer is used to perform dependency analysis on the construction worker behavior analysis data set and construct a dependency weight matrix. It should be noted that in the process of performing dependency analysis on the construction worker behavior analysis data set, the mutual dependency of each data in the construction worker behavior data can be analyzed, such as the correspondence between the daytime working period and the construction address, to provide prior knowledge for subsequent behavior prediction; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set based on the dependency weight matrix and construct predicted construction worker behavior data. It should be noted that in the dependency With the assistance of the weight matrix, it is possible to combine the prior dependencies of the behavioral data items to improve the accuracy of the behavioral data prediction; the feature enhancement layer is used to enhance the features of the construction worker behavior data obtained at the current monitoring time point based on the predicted construction worker behavior data, and construct the construction worker behavior data to be analyzed. In the process of feature enhancement, it is possible to enhance the abnormal data items of the construction worker behavior data obtained at the current monitoring time point with the assistance of the predicted construction worker behavior data, and improve the accuracy of the subsequent behavior data analysis; the behavior data analysis layer is established based on the BP neural network, and is used to analyze the construction worker behavior data to be analyzed and generate access behavior threat labels; the access behavior threat label output layer is used to output access behavior threat labels.

[0054] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A construction worker management method, characterized in that: include: Whenever a construction worker performs an access behavior, obtaining construction worker behavior data of the construction worker; The N construction worker behavior data before the current monitoring time point are combined into a construction worker behavior analysis data set, where N is the total number of construction worker behavior data in the construction worker behavior analysis data set. The construction worker behavior analysis data set and the current construction worker behavior data are then sent to the access behavior threat analysis model for processing, and an access behavior threat label is output. The access behavior threat label includes threat and no threat. If the output access behavior threat label is threat, all access behaviors of the construction worker are interrupted to avoid further data leakage, and a dynamic password for executing access behavior is randomly generated for the construction worker; if the output access behavior threat label is not threat, no operation is performed to achieve management of the construction worker; The access behavior threat analysis model includes a dependency weight matrix construction layer, a behavior data prediction layer, a feature enhancement layer, a behavior data analysis layer, and an access behavior threat label output layer; The dependency weight matrix construction layer is used to perform dependency analysis on the construction worker behavior analysis data set and construct a dependency weight matrix; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set based on the dependency weight matrix and construct predicted construction worker behavior data; The feature enhancement layer is used to enhance the features of the construction worker behavior data obtained at the current monitoring time point based on the predicted construction worker behavior data, and construct the construction worker behavior data to be analyzed; the behavior data analysis layer is used to analyze the construction worker behavior data to be analyzed and generate access behavior threat labels; The access behavior threat label output layer is used to output the access behavior threat label; The dependency analysis of the construction worker behavior analysis dataset is performed through the dependency weight matrix construction layer to construct a dependency weight matrix, which specifically includes the following steps: Construct N-1 sets of construction worker behavior analysis differential data. The construction method of the construction worker behavior analysis differential data is as follows: perform a differential operation on the n+1th construction worker behavior data in the construction worker behavior analysis data set and the nth construction worker behavior data in the construction worker behavior analysis data set to construct the nth construction worker behavior analysis differential data, n=1, 2, 3, ..., N-1; Traverse N-1 construction worker behavior analysis differential data, and for the selected i-th construction worker behavior analysis differential data, i=1, 2, 3, ..., N-2, perform the following operations: multiply the i-th construction worker behavior analysis differential data with the differential key matrix to construct the construction worker behavior analysis key matrix K, multiply the i+1-th construction worker behavior analysis differential data with the differential query key matrix to construct the construction worker behavior analysis query matrix Q, and calculate the differential attention weight matrix A i =softmax(QK T ), where T is the transposition operation; until N-1 construction workers' behavior analysis differential data are all traversed, the N-1 differential attention weight matrix A i Perform an averaging operation to construct a dependency weight matrix.

2. A construction worker management method according to claim 1, characterized in that: In the behavior data prediction layer, the behavior data of the construction worker behavior analysis dataset is predicted based on the dependency weight matrix to construct the predicted construction worker behavior data, which specifically includes the following steps: The behavior data prediction layer is built based on the LSTM model. The difference from the traditional LSTM model is that every time a hidden vector is generated in the behavior data prediction layer, the hidden vector is updated with the dependency weight matrix. The behavior data prediction layer is used to predict the behavior data of the construction workers' behavior analysis dataset and construct the predicted construction workers' behavior data. The hidden vector is updated with the dependency weight matrix, which specifically includes the following steps: multiplying the hidden vector with the hidden key matrix and the hidden value matrix respectively to construct the behavior data hidden key matrix H and the behavior data hidden value matrix G; multiplying the dependency weight matrix with the hidden query key matrix to construct the behavior data hidden query matrix R; and calculating the hidden attention weight matrix B = softmax (RH T ), and then multiply the behavior data hidden value matrix G with the hidden attention weight matrix B to update the hidden vector.

3. A construction worker management method according to claim 2, characterized in that: The feature enhancement layer is used to enhance the features of the construction worker behavior data obtained at the current monitoring time point, and the construction worker behavior data to be analyzed is constructed, which specifically includes the following steps: Performing a differential operation on the construction worker behavior data obtained at the current monitoring time point and the predicted construction worker behavior data to construct the construction worker behavior differential data, and enhancing the features of the construction worker behavior differential data through the dependency weight matrix to construct the construction worker behavior data to be analyzed; The feature enhancement of construction workers' behavior differential data is performed by relying on the weight matrix, which specifically includes the following steps: multiplying the construction workers' behavior differential data with the enhancement key matrix and the enhancement value matrix respectively, constructing the behavior data enhancement key matrix X and the behavior data enhancement value matrix Y, multiplying the dependency weight matrix with the enhancement query key matrix, constructing the behavior data enhancement query matrix E, and calculating the enhancement attention weight matrix C = softmax (EX T ), and then multiply the behavior data reinforcement value matrix Y and the reinforcement attention weight matrix C to enhance the features of the construction workers' behavior differential data and construct the construction workers' behavior data to be analyzed.

4. A construction worker management method according to claim 3, characterized in that: Training the access behavior threat analysis model includes the following steps: Obtain several positive training samples and negative training samples for access behavior threat analysis, wherein the positive training samples for access behavior threat analysis include N+1 pieces of construction worker behavior data arranged in chronological order without data leakage, and the negative training samples for access behavior threat analysis include N+1 pieces of construction worker behavior data arranged in chronological order with data leakage, and the last piece of construction worker behavior data in the negative training samples for access behavior threat analysis is obtained when data leakage occurs, annotate the positive training samples for access behavior threat analysis with non-threatening access behavior threat labels, annotate the negative training samples for access behavior threat analysis with threatening access behavior threat labels, and All labeled access behavior threat analysis positive training samples and access behavior threat analysis negative training samples constitute the access behavior threat analysis training set. The access behavior threat analysis model with parameter initialization is trained through the access behavior threat analysis training set, and the loss value is calculated. Based on the loss value, the parameters in the access behavior threat analysis model are adjusted through the back propagation algorithm to obtain the accuracy of the access behavior threat analysis model, and determine whether the accuracy of the access behavior threat analysis model is higher than the preset threshold. If the accuracy of the access behavior threat analysis model is higher than the preset threshold, the trained access behavior threat analysis model is output; otherwise, the access behavior threat analysis model continues to be trained through the access behavior threat analysis training set.

5. A construction worker management method according to claim 4, characterized in that: The behavior data analysis layer is built based on BP neural network.

6. A construction worker management method according to claim 5, characterized in that: The hyperparameters within the access behavior threat analysis model are simulated and set using a swarm optimization algorithm.

7. A construction worker management system, characterized in that: The system applies a construction worker management method according to any one of claims 1 to 6, including: A construction worker behavior data acquisition module, used for acquiring construction worker behavior data of the construction worker whenever the construction worker performs an access behavior; The access behavior threat analysis module is used to form a construction worker behavior analysis data set with N pieces of construction worker behavior data before the current monitoring time point, where N is the total number of construction worker behavior data in the construction worker behavior analysis data set, and then send the construction worker behavior analysis data set and the current construction worker behavior data to the access behavior threat analysis model for processing, and output an access behavior threat label, which includes threat and no threat. If the output access behavior threat label is threat, all access behaviors of the construction worker are interrupted to avoid further leakage of data, and a dynamic password for executing access behavior is randomly generated for the construction worker; if the output access behavior threat label is not threat, no operation is performed to achieve management of the construction worker; The access behavior threat analysis model includes a dependency weight matrix construction layer, a behavior data prediction layer, a feature enhancement layer, a behavior data analysis layer and an access behavior threat label output layer; wherein the dependency weight matrix construction layer is used to perform dependency analysis on the construction worker behavior analysis data set and construct a dependency weight matrix; the behavior data prediction layer is used to perform behavior data prediction on the construction worker behavior analysis data set based on the dependency weight matrix and construct predicted construction worker behavior data; the feature enhancement layer is used to perform feature enhancement on the construction worker behavior data acquired at the current monitoring time point based on the predicted construction worker behavior data and construct the construction worker behavior data to be analyzed; the behavior data analysis layer is used to analyze the construction worker behavior data to be analyzed and generate access behavior threat labels; the access behavior threat label output layer is used to output access behavior threat labels.

Images