A joint mapping-based fusion portrait construction and recognition method and system

Abstract

This invention relates to the field of construction safety training and data analysis technology, and provides a method and system for constructing and recognizing fusion profiles based on joint mapping. The method acquires structured input data and parameter data corresponding to construction personnel safety training. The structured input data includes a labeled master table, static attribute feature vectors, and dynamic training behavior feature vectors. The parameter data includes joint mapping rules and a pattern recognition rule base. Static and dynamic profiles are constructed separately to obtain static and dynamic profile category labels, and training attention priorities and dynamic behavior levels are configured for each. These two labels are combined according to the joint mapping rules to generate a fusion profile type identifier. Finally, based on the training attention priority and dynamic behavior level, the pattern recognition rule base is called to determine the pattern type. This method can represent and uniformly identify different characteristics, improving the targeting of training management and the standardization of output results.

Description

Technical Field

[0001] This invention relates to the field of construction safety training and data analysis technology, specifically to a method and system for constructing and recognizing fusion profiles based on joint mapping. Background Technology

[0002] Construction safety training management typically relies on statistics based on indicators such as attendance, exam scores, and training duration. While these metrics reflect overall implementation, they struggle to identify individual differences in basic attributes, risk exposure, and recent training behavior. Some solutions incorporate profiling or cluster analysis, but these often focus only on static attributes or single dimensions of behavior data from a single period, lacking hierarchical modeling and joint judgment mechanisms for both types of features. Furthermore, relying on human experience to identify high-risk or low-participation individuals suffers from inconsistent criteria, low efficiency, and insufficient reusability.

[0003] With the accumulation of data from training and on-site management platforms, information such as construction workers' job types, positions, work experience, training participation records, learning engagement, and test performance can be continuously collected. As the level of information management improves, the amount of data generated during construction safety training is gradually increasing. How to form more accurate and standardized management identification results based on this data to improve management efficiency has become a technical problem that needs to be solved in this field. Summary of the Invention

[0004] This invention aims to address the problem of low data utilization efficiency in the construction safety training process in existing technologies, and provides a method and system for constructing and recognizing fusion profiles based on joint mapping that can improve data utilization efficiency.

[0005] To achieve the above objectives, the first aspect of this application provides a method for constructing and recognizing fusion profiles based on joint mapping, comprising:

[0006] S100. Obtain structured input data and parameter data corresponding to the safety training of construction personnel. The structured input data includes a tagged main table with the unique identifier of the construction personnel as the primary key, as well as the static attribute feature vector and dynamic training behavior feature vector corresponding to the construction personnel. The parameter data includes at least joint mapping rules and a pattern recognition rule base. S200. Construct static profiles for each construction worker based on static attribute feature vectors to obtain static profile category labels for each construction worker; classify and statistically analyze the static attribute features corresponding to the construction workers according to the static profile category labels; and configure training and attention priorities for each static profile category label based on the classification and statistical results. S300. Based on the dynamic training behavior feature vector, construct a dynamic profile for each construction worker and obtain the dynamic profile category label for each construction worker; classify and statistically analyze the dynamic training behavior features corresponding to the construction workers according to the dynamic profile category label, and configure dynamic behavior level for each dynamic profile category label based on the classification and statistical results. S400. For the same construction worker, combine the corresponding static portrait category label and dynamic portrait category label according to the joint mapping rule to obtain the joint mapping result, and generate a fused portrait type identifier based on the joint mapping result. The joint mapping result includes at least the training attention priority corresponding to the static portrait category label and the dynamic behavior level corresponding to the dynamic portrait category label. S500. For each construction worker, based on the joint mapping result, the pattern recognition rule base is called to perform pattern determination on the fused profile type identifier to obtain the pattern type corresponding to each construction worker. The pattern type is used to convert the fused profile type identifier into intervention suggestions for the corresponding construction worker.

[0007] The second aspect of this application provides a fusion profile construction and recognition system based on joint mapping, including: The data acquisition module is used to acquire structured input data and parameter data corresponding to the safety training of construction personnel. The structured input data includes a tagged main table with the unique identifier of the construction personnel as the primary key, as well as static attribute feature vectors and dynamic training behavior feature vectors corresponding to the construction personnel. The parameter data includes at least joint mapping rules and a pattern recognition rule base. The static profile building module is used to build static profiles for each construction worker based on static attribute feature vectors, and obtain the corresponding static profile category label for each construction worker; it classifies and statistically analyzes the static attribute features corresponding to the construction workers according to the static profile category label, and configures the training attention priority for each static profile category label based on the classification and statistical results; The dynamic profile construction module constructs dynamic profiles for each construction worker based on dynamic training behavior feature vectors, obtaining dynamic profile category labels for each construction worker; it then classifies and statistically analyzes the dynamic training behavior features of the construction workers according to the dynamic profile category labels, and configures dynamic behavior levels for each dynamic profile category label based on the classification and statistical results. The joint mapping module combines the static profile category label and the dynamic profile category label of the same construction worker according to the joint mapping rules to obtain the joint mapping result, and generates a fused profile type identifier based on the joint mapping result. The joint mapping result includes at least the training attention priority corresponding to the static profile category label and the dynamic behavior level corresponding to the dynamic profile category label. The pattern recognition module, for each construction worker, calls the pattern recognition rule base to determine the pattern of the fused profile type identifier based on the joint mapping result, and obtains the pattern type corresponding to each construction worker. The pattern type is used to convert the fused profile type identifier into intervention suggestions for the corresponding construction worker.

[0008] Compared with the prior art, this application has the following beneficial effects: This application constructs corresponding profile categories based on static attribute feature vectors and dynamic training behavior feature vectors, respectively. This allows for the separate representation of relatively stable attribute characteristics and phased training behavior characteristics of construction workers, preventing the mixing of different types of information in the same processing process. Furthermore, the static and dynamic profile category labels are combined according to joint mapping rules to form a fused profile type identifier. This enables previously independent static and dynamic data to form associated results around the same construction worker, thereby improving data utilization. By combining training focus priorities and dynamic behavior levels, and calling a pattern recognition rule base for pattern determination, the previously experience-dependent recognition process is transformed into a rule-based and repeatable determination process. This allows the data on the corresponding construction workers to be used efficiently beyond simply reflecting their attributes and behavioral states, achieving unified identification of training-related results, improving the targeting of training management, and ultimately increasing management efficiency. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A schematic flowchart of a method according to an embodiment of this application is shown. Figure 2 This illustration schematically shows a general processing framework for fusion portrait construction and recognition according to an embodiment of this application; Figure 3 This illustration schematically shows a static image construction process according to an embodiment of the present application; Figure 4 This illustration schematically shows a dynamic portrait construction process according to an embodiment of this application; Figure 5 This illustration schematically shows the generation of joint mapping results and the construction of joint frequency matrix and joint proportion matrix according to embodiments of this application; Figure 6The illustration shows a pattern recognition diagram based on training attention priority and dynamic behavior level according to an embodiment of this application; Figure 7 This illustration schematically shows a diagram of differentiated static group identification and priority ranking processing according to an embodiment of this application; Figure 8 The illustration schematically shows normalized feature radar charts of different static image categories on each static attribute dimension according to embodiments of this application. Detailed Implementation

[0011] To facilitate understanding of this application, the following description will be more comprehensive and detailed in conjunction with the accompanying drawings and preferred embodiments, but the scope of protection of this application is not limited to the following specific embodiments.

[0012] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art. The technical terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of this application.

[0013] The terms "comprising" and "having," and any variations thereof, used in this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or apparatus. Furthermore, the use of "and / or" in this application indicates at least one of the connected objects, such as A and / or B and / or C, representing seven possibilities: including A alone, B alone, C alone, and the presence of both A and B, both B and C, both A and C, and the presence of A, B, and C.

[0014] In this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0015] The following is a detailed description of the fusion profile construction and recognition method and system based on joint mapping provided by the present invention.

[0016] See Figure 1 , Figure 1 This is a flowchart illustrating a method for constructing and recognizing fused profiles based on joint mapping, provided in an embodiment of the present invention.

[0017] like Figure 1As shown, the present invention provides a method for constructing and recognizing fused profiles based on joint mapping, which includes the following steps: S100. Obtain the structured input data and parameter data corresponding to the safety training of construction personnel. The structured input data includes a tagged main table with the unique identifier of the construction personnel as the primary key, the static attribute feature vector and the dynamic training behavior feature vector corresponding to each construction personnel. The parameter data includes joint mapping rules, pattern recognition rule base, collection window identifier and version traceability information. S200. Construct static profiles for each construction worker based on static attribute feature vectors to obtain static profile category labels for each construction worker; classify and statistically analyze the construction workers according to the static profile category labels, and configure training and attention priorities for each static profile category label based on the statistical results of the static attribute features corresponding to each static profile category label. S300. Based on the dynamic training behavior feature vector, construct a dynamic profile for each construction worker and obtain the dynamic profile category label for each construction worker; classify and statistically analyze the construction workers according to the dynamic profile category label, and configure dynamic behavior level for each dynamic profile category label based on the statistical results of the dynamic training behavior features corresponding to each dynamic profile category label. S400. For the same construction worker, combine their static profile category label and dynamic profile category label according to the joint mapping rule to generate a joint mapping result, and generate a fused profile type identifier based on the joint mapping result, and associate it with the corresponding training attention priority and dynamic behavior level. S500: Based on the training attention priority and dynamic behavior level associated in the joint mapping results, the pattern recognition rule base is called to determine the pattern of the fusion profile type identifier to obtain the corresponding pattern type. The pattern type includes at least the coupled pattern and the conflicting pattern. The pattern type is used to further transform the fusion profile type identifier into the corresponding intervention suggestion category or intervention suggestion code. It can also be combined with the joint proportion, conditional proportion or priority score to add additional tags or sort the intervention priority of the corresponding fusion profile type to generate structured output results.

[0018] In this step, the pattern recognition rule base is used to transform the correspondence between training focus priorities and dynamic behavior levels into executable judgment rules. Coupled patterns represent the matching relationship between static training focus needs and dynamic training participation behavior states. Conflicting patterns represent the mismatch between static training focus needs and dynamic training participation behavior states. During pattern determination, the training focus priority and dynamic behavior level corresponding to the fused profile type identifier are first read, and then the corresponding pattern type is output according to the pattern recognition rule base. Through this step, the category combination results can be transformed into recognition results with management significance.

[0019] In this embodiment, static and dynamic profiles of construction workers are constructed separately, and then these two profiles are jointly mapped to form a fused profile type. Pattern determination is then performed based on the correspondence between training focus priority and dynamic behavior level, ultimately outputting individual-level and group-level distribution results. This enables the unified construction, joint identification, and structured output of construction worker training profiles, improving the collaborative analysis capabilities of static and dynamic information, enhancing the accuracy of key group identification, and improving the traceability and applicability of the results.

[0020] In a further embodiment, it also includes: S600, generating a structured output result based on the unique identifier of the construction worker, the static portrait category label, the dynamic portrait category label, the fused portrait type identifier, the mode type, the acquisition window identifier, and the version traceability information.

[0021] The structured output results include individual-level result tables and group-level joint profile distribution results. The individual-level result tables include unique identifiers for construction workers, static profile category labels, dynamic profile category labels, fused profile type identifiers, mode types, acquisition window identifiers, and version traceability information. The group-level joint profile distribution results include at least a joint frequency matrix, a joint proportion matrix, and a list of fused profile type identifiers corresponding to different mode types.

[0022] In one embodiment, the static attribute feature vector includes at least one or more features of the construction personnel, such as age, education level, job type, entry status, mechanized operation status, high-risk operation status, and collaborative operation status. The dynamic training behavior feature vector includes at least one or more features of the training duration, course completion rate, learning engagement, and learning points.

[0023] Specifically, this embodiment further defines the composition of the static attribute feature vector and the dynamic training behavior feature vector. Features such as age, education level, job type, entry status, mechanized operation status, high-risk operation status, and collaborative operation status can be used to reflect differences in basic attributes, work environment, and operational risks among construction personnel, thereby supporting static profile construction and training focus priority configuration. Features such as learning duration, course completion rate, learning engagement, and learning points can be used to reflect the training status and behavioral level of construction personnel within the current data collection window, thereby supporting dynamic profile construction and dynamic behavior level configuration. The above features can be selected individually or in combination, and can be expanded or reduced according to actual business scenarios. By clearly defining the input feature content, the data foundation for static and dynamic profile construction is improved, enhancing the feasibility of this invention's implementation.

[0024] The specific implementation of the above embodiments can be carried out step by step as follows: In S100, a unique identifier for each construction worker is used to associate the static attribute data, dynamic training behavior data, and subsequent output results of the same worker. A tagged master table carries basic records and subsequently generated static profile category tags, dynamic profile category tags, fused profile type identifiers, and pattern types. Static attribute feature vectors reflect the relatively stable attributes of construction workers, while dynamic training behavior feature vectors reflect their training behavior within the current data collection window. Joint mapping rules at least include pairing the static profile category tags and dynamic profile category tags of the same construction worker within the same data collection window according to a preset combination format. A pattern recognition rule base is used to define the relationship between training focus priority and dynamic behavior level. Data collection window identifiers distinguish different statistical periods, and version tracking information records rule versions, model versions, or parameter versions for result verification and historical tracking.

[0025] In one example, the tagged main table uses OneID (a unique identifier for construction workers) as the primary key. The static attribute feature vector of the construction workers is denoted as . The training behavior feature vector is denoted as .

[0026] Optionally, a fused feature vector can be constructed: Optional fusion feature representations used for unified storage, retrieval, or statistical analysis are not required inputs for joint mapping and pattern recognition.

[0027] In addition, the S100 can load other configurable parameters, including static image construction parameters, dynamic image construction parameters, pattern recognition rule base parameters, priority mapping parameters, and threshold parameters. Among them, static image construction parameters may include the range of candidate cluster numbers and initialization method, while dynamic image construction parameters may include the range of dynamic category numbers and clustering-related parameters.

[0028] like Figure 3 As shown, in step S200, static profiles of construction workers are constructed based on static attribute feature vectors to obtain static profile category labels. This step is used to extract group features from the stable attribute differences of construction workers and to classify construction workers with similar attributes into the same static profile category. The static profile category label is used to characterize the static group type to which the construction workers belong. Subsequently, classification statistics are performed according to the static profile category labels, and training attention priorities are configured based on the statistical results of the static attribute features corresponding to each static category.

[0029] Specifically, missing values ​​can be processed first on the static attribute feature vector, and dimensionality reduction can be performed optionally. The dimensionality-reduced representation can be written as:

[0030] in, It can be used for linear dimensionality reduction methods such as truncated SVD.

[0031] Based on the processed feature vector or Perform clustering or grouping to obtain the first The static portrait category tag for each construction worker :

[0032] in, This is a collection of static image categories. The number of static image categories can be determined based on one or more of the following: intra-class error index, contour coefficient, and information criteria.

[0033] Optionally, after obtaining the static profile category labels, statistics can be compiled on one or more of the following for each static category: age structure, education structure, job type composition, entry status, proportion of high-risk operations, proportion of mechanized operations, and proportion of collaborative operations. Based on the statistical results, the training focus priority can be configured to one of three levels: high, medium, or low, or configured to a numerical level corresponding to each of the three levels. For any static category... Its training focuses on priorities It can be represented as:

[0034] The higher the value, the higher the priority of training. Semantic names can also be configured for static categories. (For example, "high-risk special operation type" and "basic job adaptability type") are used for result display and management understanding, and the semantic name of the static category is not a necessary input for pattern recognition.

[0035] In one example, training focus priority can be determined using a three-tier system of high, medium, and low. If the business presentation layer requires more detailed expression, "medium-high" can be set as the presentation level. During the pattern recognition rule execution process, "medium-high" can be merged into the "high" level according to preset rules to ensure consistency in rule judgment.

[0036] Priority assignment can be done according to rules, for example: Static categories with a high proportion of high-risk operations are assigned medium to high priority. Static categories with a high proportion of new entrants are assigned medium to high priority. Static categories with a high proportion of large-scale machinery and equipment operations are assigned medium to high priority. Static categories with low risk exposure and a high proportion of basic operations are assigned medium priority.

[0037] The rules can be preset and can be parameterized according to project type, job structure or management strategy.

[0038] It should be noted that the category semantic names of static images are only used for result display, management understanding, or interpretability enhancement, and are not required input for pattern recognition; pattern recognition is performed at least based on category identifiers and their level mapping results.

[0039] like Figure 4 As shown, in S300, dynamic profiles of construction workers are constructed based on dynamic training behavior feature vectors to obtain dynamic profile category labels. This step reflects the training participation status and behavioral differences of construction workers within the current data collection window, and classifies construction workers with similar behavioral characteristics into the same dynamic profile category. Classification statistics are performed according to the dynamic profile category labels, and dynamic behavior levels are configured based on the statistical results of dynamic training behavior features corresponding to each dynamic category.

[0040] Specifically, dynamic training behavior feature vector It is composed of one or more of the following: learning duration, course completion rate, learning engagement, and learning points. After handling missing values ​​and standardization, further dimensionality reduction can be performed, followed by clustering or grouping methods to construct a dynamic profile, resulting in the [number]th [character / item]. Dynamic portrait category tags for each construction worker :

[0041] in, A collection of dynamic portrait categories. This represents the number of dynamic image categories. The number of dynamic image categories can be determined based on one or more of the following: intra-category error index, profile coefficient, or information criteria.

[0042] For any dynamic category Dynamic behavior levels can be configured based on statistical results of behavioral indicators such as learning duration, course completion rate, learning engagement, and learning points.

[0043] Higher numerical values ​​indicate more proactive behavior and greater learning engagement and effectiveness. The dynamic behavior level corresponding to each dynamic profile category label can be configured as one of high, medium, or low levels, or as a numerical level corresponding one-to-one with high, medium, and low levels. Optionally, semantic names can also be configured for dynamic categories. (For example, "active learning type", "exam-driven type", "low-participation type", etc.), used for result display and management understanding. The semantic name of the dynamic category is not a necessary input for pattern recognition.

[0044] like Figure 5 As shown in S400, the static and dynamic image category labels corresponding to the same construction worker under the same acquisition window are jointly mapped to obtain the joint mapping result, and a fused image type identifier is generated based on the joint mapping result. For the first... The joint mapping result for the number of construction workers can be expressed as:

[0045] in, Indicates the first The combined result of static portrait category labels and dynamic portrait category labels for each construction worker.

[0046] Correspondingly, the merged image type identifier can be represented as:

[0047] in, Used to uniquely represent the static and dynamic combined category to which construction workers belong, for example In addition to retaining the static and dynamic profile category labels, the joint mapping results can also be associated with training focus priorities. and dynamic behavior levels This serves as the input basis for subsequent pattern determination.

[0048] In one embodiment, a joint frequency matrix and a joint percentage matrix are constructed based on the joint mapping results of all construction workers. The joint frequency matrix is ​​used to represent the number of construction workers corresponding to each static portrait category label and each dynamic portrait category label, and the joint percentage matrix is ​​used to represent the percentage of each corresponding construction worker in the total number of construction workers.

[0049] Specifically, let the total number of samples be... Then the elements in the joint frequency matrix can be defined as:

[0050] The elements in the joint proportion matrix can be defined as:

[0051] in, Indicates the number of elements in the set. This indicates that it belongs to the static category at the same time. and dynamic categories The number of construction workers This indicates the corresponding percentage of the total.

[0052] Optionally, the dynamic distribution proportion within each static category can be further calculated to characterize the internal structural differences of the same static group under different behavioral states. This proportion can be expressed as... :

[0053] in, Indicates in static categories Internally, it is in dynamic categories The corresponding percentage of conditions. Indicates belonging to the static category The total number of construction workers, Indicates the number of dynamic image categories. This is the summation index used when summing all dynamic image categories. Represents the static image category index; This represents the index of dynamic image categories.

[0054] In S500, based on the training attention priority and dynamic behavior level associated in the joint mapping results, the pattern recognition rule base is invoked to determine the pattern of the fused profile type identifier, thereby obtaining the corresponding pattern type. Pattern types include at least coupled patterns and conflict patterns. The pattern recognition rule base is used to transform the correspondence between training focus priorities and dynamic behavior levels into executable decision rules.

[0055] like Figure 6 As shown, specifically, it can be based on static categories. Corresponding training priorities With dynamic categories Corresponding dynamic behavior level Identification of merged portrait types (As shown in the picture) ) Perform pattern determination.

[0056] In one example, the pattern recognition rule base may contain at least the following rules (examples), as shown in Table 1: Rule C1 (Positive Coupling): When training focuses on priority High, and dynamic behavior level When it is high, determine This is a forward coupling mode.

[0057] Rule C2 (Coupling to be Strengthened): When training focuses on priorities High, and dynamic behavior level When it is in the middle, determine This is a coupling mode to be strengthened.

[0058] Rule X1 (Key Conflict): When training focuses on priority High, and dynamic behavior level When it is low, determine This is a key conflict mode.

[0059] Rule C3 (Positive Coupling): When training focuses on priority Medium or low, and dynamic behavior level When it is high, determine This is an active coupling mode.

[0060] Rule C4 (General Coupling): When training focuses on priority Medium or low, and dynamic behavior level When the condition is in the middle, it is determined to be a general coupling mode.

[0061] Rule X2 (General Conflict): When training focuses on priority Medium or low, and dynamic behavior level When it is low, determine This is the standard conflict mode.

[0062] The high, medium, and low levels mentioned above can be implemented using enumerated values ​​or numerical thresholds, and can be configured according to project type, job structure, or management strategy.

[0063] This step transforms the combined results of static and dynamic categories into identification results with managerial significance. Table 1 shows an extended example mapping, which may include intermediate state mappings in addition to rules C1–X2.

[0064] Table 1 is a pattern recognition rule mapping table.

[0065] Note: Pattern recognition rules can be configured through a rule base; the table is for illustrative purposes only. In practical applications, more levels, risk weights, or scenario-based judgment conditions can be introduced.

[0066] In one embodiment, after determining the completion mode based on training focus priority and dynamic behavior level, the mode type is further labeled or prioritized for intervention based on the joint proportion matrix constructed from the joint mapping results of all construction personnel, or the conditional proportion obtained by normalizing the number of construction personnel corresponding to different dynamic portrait category labels within the same static portrait category label.

[0067] Specifically, after obtaining the pattern type corresponding to each integrated profile type based on the training focus priority and dynamic behavior level, the joint proportion corresponding to each integrated profile type in the joint proportion matrix is ​​obtained, as well as the condition proportion corresponding to each dynamic profile category label within the same static profile category label.

[0068] When the total condition percentage of the same static profile category label in the dynamic profile category labels corresponding to the high dynamic behavior level is not less than the first threshold, and the total condition percentage of the same static profile category label in the dynamic profile category labels corresponding to the low dynamic behavior level is not less than the second threshold, the static profile category label is marked as a differentiated static group; for the fused profile type belonging to the static profile category label, a differentiated attention mark can be added on the basis of the original pattern type.

[0069] Furthermore, the intervention priority of key conflict patterns and coupling patterns to be strengthened can be ranked by combining the joint proportion of the fusion profile types; the higher the joint proportion, the greater the coverage of the fusion profile type in the overall sample, and the higher its intervention priority.

[0070] For example, when static profile category S3 is identified as a differentiated static group and fused profile type S3-D3 has been determined as a key conflict pattern, a differentiation attention marker is added on the basis of the key conflict pattern; and combined with the joint proportion corresponding to S3-D3, the ranking position of this type of object in the intervention task is improved.

[0071] In S600, structured output results are generated based on the unique identifier of construction personnel, static portrait category labels, dynamic portrait category labels, fused portrait type identifier, mode type, acquisition window identifier, and version traceability information. The structured output results include at least individual-level result tables and group-level joint distribution results.

[0072] The individual-level results table uses OneID as the primary key and includes at least the static portrait category label. Dynamic portrait category tags Integration of portrait type identifiers Mode type Collection window identifier and version tracking information.

[0073] The group-level joint distribution results should include at least the joint frequency matrix. Joint percentage matrix And a list of fused profile type identifiers corresponding to different pattern types. Optionally, the group-level joint distribution results may also include the dynamic distribution proportion within each static category. .

[0074] like Figure 7 As shown, in an optional embodiment, in order to identify behavioral differentiation phenomena within the same static category, differentiation detection can be performed on the distribution of each static category in the dynamic space.

[0075] The specific steps include: based on the joint proportion matrix, within each static portrait category label, the number of construction workers corresponding to different dynamic portrait category labels is normalized to obtain the condition proportion; when the total condition proportion of the same static portrait category label in the dynamic portrait category labels corresponding to high dynamic behavior levels is not less than the first threshold, and the total condition proportion in the dynamic portrait category labels corresponding to low dynamic behavior levels is not less than the second threshold, the construction worker group corresponding to the static portrait category label is marked as a differentiated static group.

[0076] For static categories In dynamic categories The proportion of conditions It can be represented as:

[0077] When the same static category simultaneously satisfies: , ; Static categories can be Marked as a differentiated static population. Among them, and This is a configurable threshold parameter. In the three types of dynamic profiling scenarios, the dynamic categories corresponding to high behavior levels and the dynamic categories corresponding to low behavior levels can be used separately for judgment.

[0078] In an optional embodiment, the fused portrait type identifier can also be prioritized for use in object sorting.

[0079] in, This represents the maximum value of the dynamic behavior level. , and All of these are configurable weight parameters. This represents the percentage of the integrated profile type identifier in the overall sample. A higher score indicates a higher intervention priority for that integrated profile type identifier in training management.

[0080] In one embodiment, the method further includes S700: repeating S200 to S500 under a preset period, and comparing the joint mapping distribution results and pattern type distribution results corresponding to different preset periods; when the distribution change exceeds a preset threshold, adjusting the profile construction parameters or pattern recognition rules, and writing the adjusted parameter version information into the version traceability information.

[0081] Specifically, the preset period can be set by day, week, month, or project stage. Within each preset period, based on the static attribute feature vector and dynamic training behavior feature vector corresponding to that period, static profile construction, dynamic profile construction, joint mapping, and pattern recognition are re-executed to obtain the fused profile result and pattern recognition result corresponding to that preset period.

[0082] Furthermore, by comparing the joint distribution matrices of different periods... Alternatively, changes in pattern type can identify variations in the structural distribution, behavioral changes, and pattern migration of the construction worker group. When the changes in the static profile category distribution, dynamic profile category distribution, joint mapping distribution, or pattern type distribution exceed a preset threshold, it indicates a decrease in the adaptability of the current profile construction parameters or pattern recognition rules to the existing data. In this case, the profile construction parameters or pattern recognition rules can be adjusted, and the adjusted version number and parameter information can be written into the version traceability information.

[0083] This embodiment enables the present invention to have the ability to dynamically update across cycles, improve the adaptability of the portrait results and pattern recognition results to changes in the actual training status, and at the same time ensure that the update process is recordable, verifiable and traceable.

[0084] The following embodiments are further descriptions based on the aforementioned S100 to S600.

[0085] In one specific embodiment, a set of static profile categories is constructed based on the static attribute feature vectors of construction workers. Several types of static images are obtained, the first... The static portrait category label for each construction worker is denoted as follows: ,satisfy:

[0086] To better illustrate the differences between categories, the following explanation uses six types of static portraits as examples, denoted as S1 to S6.

[0087] The processing flow corresponding to this embodiment is as follows: Figure 3 As shown, the process includes steps such as static attribute feature vector input, optional dimensionality reduction, static category generation, attribute structure statistics, static profile semantic naming, and training attention priority assignment.

[0088] By combining the distribution differences of each category in indicators such as age structure, education level, proportion of high-risk jobs, proportion of mechanized jobs, proportion of collaborative jobs, and proportion of basic jobs, the static profiles are semantically named. Radar charts can be used to display the normalized characteristics of S1 to S6 in dimensions such as the proportion of those under 35 years old, those over 45 years old, those with higher education levels, those in basic jobs, those in collaborative jobs, those in high-risk jobs, and those in mechanized jobs. Figure 8 As shown. Figure 8 This is used to characterize the static attribute structure features of different static portrait categories, and provides a basis for subsequent semantic naming of static portraits and assignment of training attention priorities. An example is shown below: S1 is the young skills enhancement category. This category has a higher proportion of personnel under 35 years old, relatively higher education levels, a greater participation in mechanized operations, and a lower proportion in high-risk special operations, reflecting the characteristics of young construction workers in terms of skills enhancement and career development.

[0089] S2 represents the older, more experienced type. This category has a higher proportion of personnel over 45 years old, generally lower educational levels, and moderate participation in high-risk and mechanical operations. They rely more on long-term construction experience to complete tasks.

[0090] S3 is a high-risk special operations category. This category has a significantly higher proportion of high-risk operations than other categories, concentrated in welding, blasting, high-altitude work, and lifting positions, making it a key focus group for safety training.

[0091] S4 is for large machinery operation. This category has a high proportion of machinery operation, involving positions operating tower cranes, hoisting equipment, mixing and paving equipment, etc., with a high degree of equipment dependence and strong job responsibilities.

[0092] S5 is a composite collaborative operation type. This category has a high proportion of collaborative operations, often involving multiple trades and multiple work areas in collaborative construction. The working environment is relatively complex, and the requirements for collaborative operation standards and training organization are relatively high.

[0093] S6 is a basic job adaptability type. This category has a high proportion of employees under 35 years old, a large proportion of whom perform basic auxiliary tasks, and a low participation rate in high-risk and mechanical operations. This is mainly reflected in a strong need for job adaptability and learning of basic safety regulations.

[0094] Based on preset rules, assign training focus priorities to the above static profiles. For any static category... Its training focus priority can be expressed as

[0095] The higher the value, the higher the priority of training focus. An example of the static profile training focus priority mapping is shown in Table 2.

[0096] Table 2 is a mapping table of attention priorities for static portrait training.

[0097] Note: Priority grading can be configured according to platform policies and is not limited to the above methods. Priority levels can be enumerated values ​​such as high, medium, and low, or represented numerically, such as 3, 2, and 1. The mapping relationships in the table are for illustrative purposes only; in actual applications, they can be parametrically adjusted according to project type, construction stage, management objectives, or historical accident risk characteristics.

[0098] In one specific embodiment, a dynamic profile category set is constructed based on the feature vector of the dynamic training behavior of construction personnel. Dynamic training participation behavior feature vectors may include indicators such as learning duration, course completion rate, learning engagement, and learning points. The dynamic profile category label for each construction worker is denoted as follows: ,satisfy:

[0099] The following uses three types of dynamic profiles as examples to illustrate their behavioral differences and hierarchical mapping relationships, denoted as D1 to D3, but this is not intended to limit the number of categories. The processing flow corresponding to this embodiment is as follows: Figure 4 As shown, the process includes steps such as dynamic training behavior feature vector input, dynamic category generation, behavior structure statistics, dynamic profile semantic naming, and dynamic behavior level assignment.

[0100] By combining the relative levels of each category in behavioral metrics, dynamic profiles can be semantically named, as shown in the following example.

[0101] D1 represents the proactive learning type. This category generally performs well in terms of learning time, course completion rate, learning engagement, and total points, demonstrating a strong willingness to learn independently and sustained engagement.

[0102] D2 is exam-driven. Course completion rates are high, and learning engagement and grades are at a moderate level, but the learning time is relatively limited, exhibiting results-oriented, exam-oriented learning characteristics.

[0103] D3 represents low participation. Overall, behavioral indicators in this category are low, indicating insufficient enthusiasm for training participation, and relatively weak learning engagement and outcome performance.

[0104] Based on preset rules, dynamic behavior levels are assigned to dynamic profiles. For any dynamic category... Its dynamic behavior level can be represented as:

[0105] The higher the value, the more positive the behavior. An example of the dynamic profile behavior level mapping is shown in Table 3.

[0106] Table 3 is a dynamic profile behavior level mapping table.

[0107] Table Explanation: 1) Behavioral levels can be represented by enumerated values ​​(high / medium / low) or numerical values ​​(3 / 2 / 1); 2) Dynamic category naming and level mapping can be redefined based on the distribution characteristics of platform behavioral indicators, and are not limited to the above examples; 3) The number of dynamic categories and their names can vary in different projects or cycles, but they can all be accessed through the joint mapping and pattern recognition process via the level mapping mechanism of this invention.

[0108] In one specific embodiment, joint mapping is performed on each construction worker to obtain the joint mapping result and the fused profile type identifier. The joint mapping result of the construction workers can be expressed as:

[0109] The corresponding merged image type identifier can be represented as:

[0110] The process of generating joint mapping and fusion profile types is as follows: Figure 5 As shown. Figure 5 This illustrates how static and dynamic portrait category labels form a joint mapping result, generate a fused portrait type identifier, and how the joint frequency matrix and joint proportion matrix are constructed.

[0111] In this embodiment, there are 6 static image categories and 3 dynamic image categories, resulting in a maximum of 18 fused image types. A joint frequency matrix is ​​constructed based on sample statistics. With joint proportion matrix This allows us to obtain the distribution differences of different static groups in the dynamic behavioral space.

[0112] By using joint distribution statistics, the distributional differences of different static groups in the dynamic behavioral space can be obtained. For example: The high proportions of S1, S3, and S5 in D2 indicate that a significant proportion of construction workers in these groups exhibit outcome-oriented training behavior. The high proportions of S2 and S4 in D1 suggest strong active learning characteristics among experienced and equipment-operating groups. The presence of S1 and S6 in D3 indicates insufficient training participation among younger or newly arrived groups. These joint distribution results can serve as the input basis for subsequent pattern recognition.

[0113] In one specific embodiment, based on the training focus priority of static profiles and the behavior level of dynamic profiles, a rule base is used to determine the pattern of the fused profile type identifier, resulting in coupled pattern, conflicting pattern, and other optional patterns. The relevant determination process is as follows: Figure 6 As shown. Figure 6This illustrates the process of determining coupling mode, conflict mode, and other optional modes for the integrated profile type identifier based on the training attention priority of static profile and the behavior level of dynamic profile through a rule base.

[0114] Specifically, based on static categories Corresponding training priorities and dynamic categories Corresponding dynamic behavior level Call the pattern recognition rule base to fuse profile types The pattern type is determined through a judgment process. The pattern type can be represented as:

[0115] Or it can be expressed as:

[0116] In this embodiment, the following rules may be adopted: When the static category has a high priority and the dynamic category has a low behavior level, it is determined to be a key conflict mode.

[0117] When the static category has high priority and the dynamic category has high behavior level, it is determined to be a positive coupling mode.

[0118] When the static category is of high priority and the dynamic category is of medium priority, it is determined to be a coupling mode that needs to be strengthened.

[0119] When the static category is of medium or low priority and the dynamic category is of low behavior level, it is determined to be a general conflict mode.

[0120] When the static category is of medium or low priority and the dynamic category is of high behavior level, it is determined to be an active coupling mode.

[0121] When the static category is of medium or low priority and the dynamic category is of medium priority, it is determined to be a general coupling mode; under the extended rule configuration, it can also be further included in the management of the category to be strengthened.

[0122] Based on the above rules, the following typical judgment results can be obtained.

[0123] For S3-D3, which is a combination of high-risk special operations and low participation, the static priority is high and the dynamic behavior level is low. It can be identified as a key conflict pattern. This type is a typical combination of high-risk positions and low training participation, and should be the priority intervention target.

[0124] For S5-D3, which is a combination of composite collaborative work type and low participation type, the static priority is high and the dynamic behavior level is low, which can be identified as a key conflict mode. This type of work is complex and the training participation is insufficient, which can easily lead to weak links in management.

[0125] For S4-D1, which is a combination of large-scale mechanical equipment operation type and active learning type, the static priority is high and the dynamic behavior level is high. It can be identified as a positive coupling mode. This type reflects the consistency between the equipment responsibility position and the active training behavior, and can be used as an experience promotion object.

[0126] For S2-D1, which is a combination of experienced and proactive learners, the static priority is medium and the dynamic behavior level is high, which can be identified as an active coupling mode. This type indicates that experienced construction workers have strong responsibility-driven and self-learning behaviors.

[0127] For S1-D2, a combination of youth skill enhancement and exam-driven learning, the static priority is medium, and the dynamic behavior level is medium. When using basic rules, it can be classified as a general coupling pattern. However, with rules designed to further strengthen the management of medium-level behaviors, it can be included in the category of coupling to be strengthened for more detailed management. Although this type has the foundation for training completion, the learning quality can still be improved by optimizing process learning constraints.

[0128] In one specific embodiment, for the same static category It can be used to calculate the proportion of conditions in each dynamic category. If the proportions of both high-behavior categories and low-behavior categories exceed the threshold (e.g.) , If the static category is identified as a differentiated static population, then the static category is marked as such. The differentiation identification and priority ranking process in this embodiment is as follows: Figure 7 As shown. Figure 7 This illustrates the process of performing conditional proportion statistics on dynamic distribution within the same static category, identifying differentiated static groups, and prioritizing conflict fusion profile type identifiers.

[0129] In this embodiment, S3 and S5 simultaneously exhibit a high proportion of both proactive learners and low-participants in the dynamic profile space, meeting the differentiation identification criteria. Therefore, they can be labeled as differentiated static groups. This type of group indicates that, under the same operational background, different individuals show significant differences in training participation behavior, which can be used to adopt differentiated management strategies.

[0130] In one specific embodiment, the present invention outputs individual-level and group-level structured results, as shown in Tables 4, 5, and 6 below, illustrating the output structure of the individual-level result table, the group-level joint distribution results, the pattern type list, and the intervention suggestion code or intervention suggestion category.

[0131] Table 4 shows the results at the individual level.

[0132] Table 5 shows the results of the group-level joint distribution.

[0133] Table 6 is a list of pattern types.

[0134] The individual-level results table may include the following fields: The system includes a unique identifier for construction workers (OneID), a static profile category label (e.g., S3), a dynamic profile category label (e.g., D3), a merged profile type identifier (e.g., S3-D3), a pattern type (e.g., key conflict pattern), a conflict priority level or score (optional), an intervention recommendation code (e.g., I01, I02, I03), an intervention recommendation category (e.g., enhanced training, process supervision, strategy maintenance and experience reuse), and a version number and data collection window identifier.

[0135] Group-level results may include the following: The joint frequency matrix and joint percentage matrix, the list of fusion profile type identifiers corresponding to key conflict modes, the list of fusion profile type identifiers corresponding to coupling modes to be strengthened, and the list of differentiated static groups are optional fields.

[0136] Depending on the pattern type and priority, this invention can also generate corresponding intervention suggestion categories or intervention suggestion codes. For example, it can output suggestions for enhanced training or process supervision for key conflict patterns; suggestions for process constraint optimization for patterns requiring enhanced coupling; and suggestions for strategy maintenance or experience reuse for positively coupled patterns.

[0137] This invention also provides a fusion profile construction and recognition system based on joint mapping, comprising: The data acquisition module is used to acquire structured input data and parameter data corresponding to the safety training of construction personnel. The structured input data includes a tagged main table with the unique identifier of the construction personnel as the primary key, as well as static attribute feature vectors and dynamic training behavior feature vectors corresponding to the construction personnel. The parameter data includes at least joint mapping rules and a pattern recognition rule base. The static profile building module is used to build static profiles of each construction worker based on static attribute feature vectors, obtain the corresponding static profile category labels for each construction worker, classify and statistically analyze the construction workers according to the static profile category labels, and configure training and attention priorities for each static profile category label based on the statistical results of the static attribute features corresponding to each static profile category label. The dynamic profile building module is used to build dynamic profiles for each construction worker based on dynamic training behavior feature vectors, and obtain the dynamic profile category label for each construction worker; classify and statistically analyze the construction workers according to the dynamic profile category label, and configure dynamic behavior levels for each dynamic profile category label based on the statistical results of dynamic training behavior features corresponding to each dynamic profile category label. The joint mapping module is used to combine the static portrait category label and the dynamic portrait category label of the same construction worker according to the joint mapping rules to obtain the joint mapping result, and generate a fused portrait type identifier based on the joint mapping result; The pattern recognition module is used to determine the pattern of each construction worker by calling the pattern recognition rule base based on the training attention priority and dynamic behavior level associated in the joint mapping result.

[0138] In a further embodiment, the joint mapping module further includes: constructing a joint frequency matrix and a joint proportion matrix based on the joint mapping results of all construction personnel, wherein the joint frequency matrix represents the number of construction personnel corresponding to each combination of static portrait category label and each combination of dynamic portrait category label, and the joint proportion matrix represents the proportion of the number of construction personnel corresponding to each combination to the total number of construction personnel.

[0139] In a further embodiment, the pattern recognition rule base includes the following rules: When the training focus is high and the dynamic behavior level is high, it is judged as a positive coupling mode; When the training focus is high and the dynamic behavior level is medium, it is determined to be a coupling mode that needs to be strengthened. When the training focus is high and the dynamic behavior level is low, it is judged as a key conflict mode; When the training focus priority is medium or low and the dynamic behavior level is high, it is judged as an active coupling mode; When the training focus is on a medium or low priority and the dynamic behavior level is low, it is judged as a general conflict pattern.

[0140] When the training focus priority is medium or low and the dynamic behavior level is medium, it is judged as a general coupling mode.

[0141] In a further embodiment, based on the joint frequency matrix or the static category internal proportion obtained by conversion based on the joint proportion matrix, the number of construction workers corresponding to different dynamic image category labels is normalized within each static image category label to obtain the conditional proportion. When the total proportion of construction workers corresponding to the same static portrait category label in the high dynamic behavior level category is not less than the first threshold, and the total proportion of conditions in the low dynamic behavior level category is not less than the second threshold, the construction worker group corresponding to the static portrait category label is marked as a differentiated static group.

[0142] The static image building module is specifically used for: Handle missing values ​​in static attribute feature vectors; The number of static portrait categories is determined based on one or more of the intra-class error index, contour coefficient, and information criteria. Based on the processed static attribute feature vectors and the determined number of static image categories, clustering or grouping methods are used to construct static images and obtain static image category labels. Construction workers were categorized and statistically analyzed according to static portrait category tags, and the static attribute characteristics corresponding to each static portrait category tag were statistically analyzed. Based on the statistical results, corresponding training attention priorities are configured for each static profile category label, and the training attention priorities are represented by high, medium, and low levels or corresponding numerical levels.

[0143] The dynamic profile building module is specifically used for: The dynamic training behavior feature vector is processed by handling missing values ​​and standardization to obtain the processed dynamic training behavior feature vector. The number of dynamic portrait categories is determined based on one or more of the intra-class error index, contour coefficient, and information criteria. Based on the processed dynamic training behavior feature vector and the determined number of dynamic profile categories, clustering or grouping methods are used to construct dynamic profiles and obtain dynamic profile category labels. Construction workers were categorized and statistically analyzed according to the dynamic profile category tags, and the dynamic training behavior characteristics corresponding to each dynamic profile category tag were statistically analyzed. Based on the statistical results, the dynamic behavior level corresponding to each dynamic profile category label is configured as high, medium, low or corresponding numerical level.

[0144] In a further embodiment, a structure output module is also included, which is used to generate structured output results based on the unique identifier of construction personnel, static portrait category label, dynamic portrait category label, fused portrait type identifier, pattern type, and differentiated static group.

[0145] The present invention also provides a readable storage medium storing a program or instructions that, when executed by a processor, implement the various processes of the above-described method embodiments and achieve the same technical effect. To avoid repetition, these will not be described again here.

[0146] Among them, readable storage media include computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks, etc.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0148] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for constructing and recognizing fused profiles based on joint mapping, characterized in that, Including the following steps: S100. Obtain structured input data and parameter data corresponding to the safety training of construction personnel. The structured input data includes a tagged main table with the unique identifier of the construction personnel as the primary key, as well as the static attribute feature vector and dynamic training behavior feature vector corresponding to the construction personnel. The parameter data includes at least joint mapping rules and a pattern recognition rule base. S200. Construct static profiles for each construction worker based on static attribute feature vectors to obtain static profile category labels for each construction worker; classify and statistically analyze the static attribute features corresponding to the construction workers according to the static profile category labels; and configure training and attention priorities for each static profile category label based on the classification and statistical results. S300. Based on the dynamic training behavior feature vector, construct a dynamic profile of each construction worker and obtain the dynamic profile category label corresponding to each construction worker. The dynamic training behavior characteristics of construction personnel are classified and statistically analyzed according to the dynamic profile category tags, and dynamic behavior levels are configured for each dynamic profile category tag based on the classification and statistical results. S400. For the same construction worker, combine the corresponding static portrait category label and dynamic portrait category label according to the joint mapping rule to obtain the joint mapping result, and generate a fused portrait type identifier based on the joint mapping result. The joint mapping result includes at least the training attention priority corresponding to the static portrait category label and the dynamic behavior level corresponding to the dynamic portrait category label. S500. For each construction worker, based on the joint mapping result, the pattern recognition rule base is called to perform pattern determination on the fused profile type identifier to obtain the pattern type corresponding to each construction worker. The pattern type is used to convert the fused profile type identifier into intervention suggestions for the corresponding construction worker.

2. The method according to claim 1, characterized in that, The pattern recognition rule base in S500 includes the following rules: When the training focus is high and the dynamic behavior level is high, it is judged as a positive coupling mode; When the training focus is high and the dynamic behavior level is medium, it is determined to be a coupling mode that needs to be strengthened. When the training focus is high and the dynamic behavior level is low, it is judged as a key conflict mode; When the training focus priority is medium or low and the dynamic behavior level is high, it is judged as an active coupling mode; When the training focus priority is medium or low and the dynamic behavior level is low, it is judged as a general conflict pattern. When the training focus priority is medium or low and the dynamic behavior level is medium, it is judged as a general coupling mode.

3. The method according to claim 2, characterized in that, After obtaining the joint mapping result in S400, the method further includes: constructing a joint frequency matrix and a joint proportion matrix based on the joint mapping result of all construction personnel, wherein the joint frequency matrix represents the number of construction personnel corresponding to each combination of static portrait category label and each dynamic portrait category label, and the joint proportion matrix represents the proportion of the number of construction personnel corresponding to each combination to the total number of construction personnel.

4. The method according to claim 3, characterized in that, Based on the joint proportion matrix, within each static portrait category label, the number of construction workers corresponding to different dynamic portrait category labels is normalized to obtain the conditional proportion. When the total proportion of conditions for construction workers corresponding to the same static portrait category label in categories with high dynamic behavior levels is not less than the first threshold, and the total proportion of conditions in categories with low dynamic behavior levels is not less than the second threshold, the construction worker group corresponding to the static portrait category label is marked as a differentiated static group.

5. The method according to claim 1, characterized in that, S200 specifically includes: Missing values ​​are processed in the static attribute feature vector to obtain the processed static attribute feature vector; Based on the processed static attribute feature vector, the number of static portrait categories is determined according to the intra-class error index and / or contour coefficient. Based on the processed static attribute feature vectors and the determined number of static image categories, clustering or grouping methods are used to construct static images and obtain static image category labels. Construction workers were categorized and statistically analyzed according to static portrait category tags, and the static attribute characteristics corresponding to each static portrait category tag were statistically analyzed. Based on the statistical results, corresponding training attention priorities are configured for each static profile category label, with the training attention priorities represented by high, medium, and low levels.

6. The method according to claim 1, characterized in that, The S300 specifically includes: The dynamic training behavior feature vector is processed by handling missing values ​​and standardization to obtain the processed dynamic training behavior feature vector. Based on the processed dynamic training behavior feature vector, the number of dynamic profile categories is determined according to one or more of the intra-class error index and contour coefficient. Based on the processed dynamic training behavior feature vector and the determined number of dynamic profile categories, clustering or grouping methods are used to construct dynamic profiles and obtain dynamic profile category labels. Construction workers were categorized and statistically analyzed according to the dynamic profile category tags, and the dynamic training behavior characteristics corresponding to each dynamic profile category tag were statistically analyzed. Based on the statistical results, the dynamic behavior levels corresponding to each dynamic profile category label are configured as high, medium, and low.

7. The method according to claim 1, characterized in that, The static attribute feature vector includes at least one or more features of the construction personnel, such as age, education level, job type, entry status, mechanized operation status, high-risk operation status, and collaborative operation status. The dynamic training behavior feature vector includes at least one or more features of the training duration, course completion rate, and learning points.

8. The method according to claim 4, characterized in that, Also includes: S600 generates structured output results based on the unique identifier of construction personnel, static portrait category tags, dynamic portrait category tags, fused portrait type identifier, pattern type, and differentiated static groups.

9. A fusion profile construction and recognition system based on joint mapping, characterized in that, include: The data acquisition module is used to acquire structured input data and parameter data corresponding to the safety training of construction personnel. The structured input data includes a tagged main table with the unique identifier of the construction personnel as the primary key, as well as static attribute feature vectors and dynamic training behavior feature vectors corresponding to the construction personnel. The parameter data includes at least joint mapping rules and a pattern recognition rule base. The static profile building module is used to build static profiles for each construction worker based on static attribute feature vectors, and obtain the corresponding static profile category label for each construction worker; it classifies and statistically analyzes the static attribute features corresponding to the construction workers according to the static profile category label, and configures the training attention priority for each static profile category label based on the classification and statistical results; The dynamic profile building module is used to build dynamic profiles for each construction worker based on dynamic training behavior feature vectors, and obtain the dynamic profile category label for each construction worker; classify and statistically analyze the dynamic training behavior features of the construction workers according to the dynamic profile category label, and configure dynamic behavior levels for each dynamic profile category label based on the classification and statistical results. The joint mapping module is used to combine the static profile category label and the dynamic profile category label of the same construction worker according to the joint mapping rules to obtain the joint mapping result, and generate a fused profile type identifier based on the joint mapping result. The joint mapping result includes at least the training attention priority corresponding to the static profile category label and the dynamic behavior level corresponding to the dynamic profile category label. The pattern recognition module is used to determine the pattern of the fused profile type identifier based on the joint mapping result for each construction worker, and obtain the pattern type corresponding to each construction worker. The pattern type is used to convert the fused profile type identifier into intervention suggestions for the corresponding construction worker.

10. The fusion portrait construction and recognition system based on joint mapping according to claim 9, characterized in that, The joint mapping module further includes: constructing a joint frequency matrix and a joint proportion matrix based on the joint mapping results of all construction personnel. The joint frequency matrix is ​​used to represent the number of construction personnel corresponding to each static portrait category label and each dynamic portrait category label, and the joint proportion matrix is ​​used to represent the proportion of each corresponding construction personnel in the total number of construction personnel.

Images