A data processing method for enterprise employee training

Abstract

The application discloses a data processing method for enterprise employee training, and relates to the technical field of data processing, which comprises the following steps: obtaining first training content, performing grid processing on the first training content to obtain a content area grid and a content grid sequence, and sending the content grid sequence to an interactive terminal; sending the first training content to a training display terminal; receiving reference operation data fed back by the interactive terminal, obtaining a behavior type combination based on a multi-dimensional behavior model, and generating a first analysis report; determining second training content based on the first analysis report and multi-dimensional analysis rules; obtaining a second analysis report according to the second training content, generating a cross-knowledge navigation and a cross-behavior mode curve based on the first analysis report and the second analysis report, and then determining third training content; and establishing a multi-dimensional behavior model to analyze employee operation data from multiple dimensions, and comprehensively displaying the behavior and concentration distribution of employees in the training process, so that the deep analysis effect of user behavior intention and behavior mode is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a data processing method for enterprise employee training. Background Technology

[0002] Corporate employee training is specialized training designed by companies based on their industry characteristics and development status. Its purpose is to improve employees' skills in various aspects, thereby promoting the overall development of the company. Currently, internal corporate training is typically conducted through on-site training or online lectures. However, neither of these methods can provide targeted training based on the specific circumstances of each employee.

[0003] Chinese invention patent application number 202210538885.5 provides a data processing method applicable to a training and assessment system. The method involves dividing training graphic and video frames into grids to obtain virtual area grids corresponding to each frame; generating a virtual grid sequence; sending the virtual grid sequence to a triggering end to obtain employee trigger operation data; the triggering end generating a trigger grid sequence based on the correspondence between the trigger area grid and the virtual area grid; and the server generating first training and assessment information and second training and assessment information based on the trigger area grid, the trigger target area, and the information collection layer, and sending these to the training and assessment end.

[0004] However, in existing technologies, generating corresponding training and assessment information based on employees' click locations and annotation operations has limitations in identifying users' specific learning intentions (such as level of understanding, points of interest, difficulties, etc.). It cannot accurately identify users' learning intentions and behavioral patterns, making it difficult to accurately track and optimize training effectiveness. Summary of the Invention

[0005] This application provides a data processing method for enterprise employee training, which solves the problem that the limitations of the identification dimensions in the prior art make it impossible to accurately track and optimize the training effect. It realizes the technical effect of accurately tracking and optimizing the training by deeply analyzing user behavior intentions and behavior patterns.

[0006] This application provides a data processing method for enterprise employee training, applied to a server, a training display terminal, and an interactive terminal, including:

[0007] S100: Obtain the personnel attribute information of the trainees to get the corresponding first training content, divide the first training content into grids to get content area grids and content grid sequences, send the content grid sequence to the interactive terminal; send the first training content to the training display terminal;

[0008] S200: Receives baseline operation data from the interactive terminal, obtains the combination of employee behavior types based on a multi-dimensional behavior model, and generates a first analysis report; the first analysis report includes behavioral intention analysis, focus analysis, and a heat map, used to evaluate training effectiveness;

[0009] S300: Determine the second training content based on the first analysis report and multidimensional analysis rules, and send the second training content to the training demonstration terminal;

[0010] S400: Obtain a second analysis report based on the second training content, generate cross-knowledge navigation and cross-behavioral pattern curves based on the first and second analysis reports, and then determine the third training content;

[0011] The benchmark operation data includes operation time, click location, press duration, and pressure value.

[0012] Furthermore, the operation time refers to the specific time point at which the employee performs the trigger operation; the click location refers to the specific coordinates of the click operation location; the press duration is the duration of a single trigger by the employee; and the pressure value is the pressure applied during a single trigger.

[0013] The behavioral intent analysis is obtained by analyzing combinations of multiple behavioral types; the focus analysis is obtained by comparing the calculated focus index with a pre-set focus threshold; the heatmap is used to display the distribution of user click locations, pressure values, and pressure durations, and records the focus index at different click locations to identify high-frequency operation areas.

[0014] Furthermore, the multidimensional analysis rules include:

[0015] Establish a content area grid-operation behavior matrix, identify inefficient area grids and obtain hot spots and blind spots, and build a three-dimensional knowledge network based on the hot spots, blind spots and corresponding behavior type combinations to generate second training content.

[0016] Specifically, a mapping table between content region grids and behavior type combinations is generated based on the content region grid-operation behavior matrix; content region grids with a focus index below the focus threshold are obtained and marked as inefficient region grids; each inefficient region grid is divided into several micro-grids, and the micro-grids are divided into hot spots and blind spots based on the three-dimensional data of each micro-grid; the three-dimensional data is calculated based on the baseline operation data of each micro-grid, including operation duration density, operation frequency, and dwell time.

[0017] Furthermore, based on the 3D data of each micromesh, the micromesh is divided into hotspot areas and blind areas, including:

[0018] A dwell time baseline value is preset, and microgrid areas with dwell time less than the dwell time baseline value are marked as blind zones; the thermal value of each microgrid is calculated based on 3D data, the thermal values ​​are arranged in descending order, the average value of the thermal values ​​is calculated, and the microgrids are marked according to the average value and the marking rules;

[0019] The marking rules are as follows: if the average value is not less than a preset basic threshold, the micro-grid areas corresponding to the top 60% of the thermal values ​​are marked as hot spots; the remaining micro-grid areas are marked as blind spots. If the average value is less than a preset basic threshold, the micro-grid areas corresponding to the top 20% of the thermal values ​​are marked as hot spots; the remaining micro-grid areas are marked as blind spots.

[0020] The thermal value is the product of the three-dimensional data.

[0021] Furthermore, a three-dimensional knowledge network is established based on hotspots and blind spots and their corresponding behavioral types to generate secondary training content, including:

[0022] The thermal features corresponding to hot spots and blind spots are extracted respectively. Attribution analysis and adjustment strategies are obtained based on the pre-established hot spot attribution rule base and blind spot attribution rule base. A three-dimensional knowledge network is obtained based on the thermal features, attribution analysis and adjustment strategies. The first training content is adjusted based on the three-dimensional knowledge network to obtain the second training content.

[0023] The process involves pre-establishing attribution rule bases for hotspot areas and blind areas. This includes acquiring historical training data, labeling attribution analysis based on the thermal characteristics corresponding to hotspot areas and blind areas respectively, setting corresponding adjustment strategies based on the attribution analysis, and analyzing the association rules between thermal characteristics and attribution analysis using a decision tree algorithm to obtain support. The hotspot area attribution rule base and blind area attribution rule base are then obtained based on the thermal characteristics, attribution analysis, adjustment strategies, and support for hotspot areas and blind areas respectively.

[0024] Furthermore, based on the first and second parsing reports, cross-knowledge navigation maps and cross-behavioral pattern curves are generated, including:

[0025] Based on the first and second analysis reports, correlation indicators were identified, and relevant training content was determined.

[0026] The content areas in the relevant training content are broken down into knowledge nodes, and the knowledge nodes are connected based on dependencies. The association weights between different knowledge nodes are calculated, and a cross-knowledge navigation graph is formed based on knowledge nodes, dependencies, and association weights.

[0027] The behavioral span path is determined based on the correlation indicators and the correlation training content, and a cross-behavioral pattern curve is established. The cross-behavioral pattern curve is a three-dimensional curve with time as the horizontal axis and operation duration density, pressure gradient and focus index as the vertical axis.

[0028] Furthermore, the third training content is determined, including: identifying the changing trend characteristics of cross-behavioral pattern curves, finding the learning behaviors that best match the changing trend characteristics from the cross-knowledge navigation map, and selecting the corresponding training content as the third training content based on the employee's current knowledge nodes.

[0029] Furthermore, the method also includes: S500: setting personalized trigger conditions based on the third training content, and setting a guided training mechanism according to the personalized trigger conditions; the guided training mechanism automatically compares the changes in employee behavior with the personalized trigger conditions, and automatically adjusts the learning path if the personalized trigger conditions are met.

[0030] Furthermore, the personalized triggering conditions are set based on the slope of the operation duration density, the rate of change of the pressure gradient, and the convergence radius of the focus index, including high-pressure acceleration, dispersion repair, steady-state transition, and contradiction intervention; the convergence radius of the focus index is the standard deviation of the focus index.

[0031] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0032] By utilizing the correspondence between operation area grids and content area grids, the baseline operation data of employees on the interactive terminal is accurately recorded and identified. A multi-dimensional behavior model is established to analyze employee operation data from multiple dimensions, resulting in behavioral intent analysis, focus analysis, and heat maps, which comprehensively display the distribution of employee behavior and focus during the training process. Based on the first analysis report, the employee training situation is judged, subsequent training plans are determined, and training quality is improved. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of a data processing method for enterprise employee training in an embodiment of the present invention. Detailed Implementation

[0034] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0036] Example 1: The technical solution of this application includes a server, a training demonstration terminal, and an interactive terminal. The server is used to process data from the training demonstration terminal and the interactive terminal. The training demonstration terminal is used by the training teacher to play multimedia training content. The interactive terminal can be a trigger pad that is wirelessly or wiredly connected to the server. The size of the trigger pad can be proportionally scaled down to the screen where the training video is played. When training is needed, each training employee can be provided with a trigger pad. Users can trigger the trigger pad. The server can collect the trigger information on the trigger pad in real time. The trigger pad interacts with the training demonstration terminal in real time during the training process.

[0037] The data processing method for enterprise employee training described in this application can be implemented by software and / or hardware devices. The implementing entity may include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment may include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment may include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this. It is understood that enterprise employee training is specialized training designed by enterprises based on their industry characteristics and development status. Its purpose is to improve the level of employees' knowledge, skills, work methods, work attitudes, and other aspects, thereby promoting the overall development of the enterprise.

[0038] like Figure 1 As shown, a data processing method for enterprise employee training is applied to a server, a training display terminal, and an interactive terminal, including:

[0039] S100: Obtain the personnel attribute information of the trainees to determine the corresponding first training content, extract the multimedia content frames in the first training content, perform grid processing on the multimedia content frames to obtain the corresponding content area grids; generate a content grid sequence according to the time of the multimedia content frames corresponding to each content area grid, and send the content grid sequence to the interactive terminal; send the first training content to the training display terminal.

[0040] In some embodiments, personnel attribute information can be, for example, the personnel's job attributes. For instance, if employee A is a manager, then corresponding management multimedia training content can be determined and provided to employee A for targeted training. This solution can also obtain the user's training permission information after login, and use this permission information to filter corresponding first training content from a preset database, then use this first training content to provide training to the relevant personnel. By setting corresponding first training content for each user, this solution can provide targeted training based on the user's job position or knowledge level, avoiding situations where training is unsuitable.

[0041] In some embodiments, a first quantity value is obtained based on the number of background pixels within the background pixel interval of the multimedia content frame, and a second quantity value is obtained based on the number of content pixels within the content pixel interval of the multimedia content frame. Content proportion information is generated based on the first and second quantity values. The length and width information of the multimedia content frame are obtained to obtain aspect ratio information. The current number of length dividers is calculated based on the content proportion information and the baseline number of length dividers. The current number of width dividers is calculated based on the current number of length dividers and the aspect ratio information. The length divider and width divider corresponding to the current number of length dividers and width dividers are selected, and the multimedia content frame is segmented to obtain virtual region grids. It is understood that this solution uses the content of the multimedia content frame to segment the multimedia content frame to obtain the content region grids corresponding to each multimedia content frame. The more content, the more corresponding content region grids; similarly, the less content, the fewer corresponding content region grids.

[0042] S200: Receive baseline operation data from the interactive terminal, the baseline operation data including operation time, click location, press duration and pressure value; analyze the baseline operation data based on a multi-dimensional behavior model to obtain the employee's behavior type combination and generate a first analysis report.

[0043] In some embodiments, the operation time refers to the specific time point at which the employee performs the trigger operation; the click location refers to the specific location coordinates of the click operation; the pressing duration (T) is the duration of a single trigger by the employee; and the pressure value (P) is the pressing force during a single trigger collected by a pressure sensing device, with a value range of [0, 1].

[0044] In some embodiments, the interactive terminal categorizes multiple operation areas into operation area grids based on the content area grids and the interactive terminal interface areas. An operation grid sequence is generated based on the correspondence between the operation area grids and the content area grids. The operation grid sequence includes the operation area grids and the corresponding time, which corresponds one-to-one with the content grid sequence. That is, each operation area grid corresponds to a content area grid at the same time. This relationship is achieved through the communication protocol and data mapping relationship between the interactive terminal and the server. When an employee performs an operation on an operation area grid on the interactive terminal, it will be mapped to the corresponding content area grid and trigger the corresponding processing logic, ensuring that the operation data of the employee on the interactive terminal can be accurately recorded and identified.

[0045] In some embodiments, the multidimensional behavior model specifically includes acquiring employee baseline operation data, including operation time, click location, press duration, and pressure value. When inputting data, the baseline operation data needs to undergo data preprocessing, which includes data cleaning and data normalization. Data cleaning is used to filter outliers (such as instantaneous clicks or accidental device touches), and data normalization standardizes the pressure values ​​of different devices, normalizing multiple data of different dimensions to the range of 0-1, ensuring all data are on the same dimension. Key features are extracted from the preprocessed data, such as the coordinates of the click location, the distribution of press duration, and the range of pressure values. Based on these features, employee behavior is categorized into four types: short press, long press, high pressure, and low pressure. Machine learning algorithms are used to learn and analyze the extracted features to identify different employee behavior patterns. Machine learning algorithms, such as decision trees, random forests, or neural networks, are well-known techniques and will not be elaborated upon here. The four basic types of behavior (short press, long press, high pressure, and low pressure) are combined and analyzed using machine learning algorithms to obtain behavior type combinations (long press - high pressure), resulting in behavioral intent analysis. Based on pre-set weighting coefficients, press duration, and pressure value, a focus index is calculated using a formula. The calculated focus index is compared with a pre-set focus threshold to obtain a focus analysis. A heatmap is generated based on the distribution of click locations, pressure values, and pressure duration to visually display employee operational behaviors and focus distribution during training. In the heatmap, different click locations record corresponding focus indices, presented through color depth or numerical labels, clearly identifying high-frequency areas of employee operations and helping trainers understand employees' focus and operational habits during training. The First Analysis Report comprehensively and deeply evaluates employee performance and training effectiveness through a combined analysis of behavioral intent analysis, focus analysis, and heatmaps.

[0046] In some embodiments, the first analysis report is an analysis and evaluation report generated based on a combination of baseline operational data and behavior types, used to evaluate the quality of employee training; the first analysis report includes behavioral intent analysis, focus analysis, and a heatmap; the behavioral intent analysis is obtained by randomly combining and analyzing multiple behavior types; the focus analysis is obtained by comparing and analyzing a calculated focus index with a pre-set focus threshold; the heatmap is used to display the distribution of user click locations, stress values, and stress durations, and records the focus index at different click locations to identify high-frequency operation areas.

[0047] In some embodiments, the focus index is calculated by comprehensively considering a preset weighting coefficient, pressing duration, and pressure value. The calculation formula is as follows:

[0048]

[0049] in, It is a focus index; and These are pre-set weighting coefficients. ; It is the pressure value; It is the logarithm of the press duration T. Specifically, and The process involves pre-determining the weighting coefficients for pressure value and pressing duration based on historical experimental data or opinions from internal training experts, obtaining the focus index under different weighting coefficients in historical experimental data, and selecting the optimal weighting coefficients based on the results of multiple experiments. Preferably, setting the weighting coefficient for pressure value to 0.43 and the weighting coefficient for pressing duration to 0.57 best reflects the employee's true focus index. The specific values ​​can be adjusted according to the actual situation.

[0050] Specifically, short press (T<500ms): quickly confirm or skip content; long press (T≥500ms): requires thought or marking key points; high pressure (P≥0.7): high focus or response to high-difficulty content; low pressure (P<0.3): distraction or low interest. Combining these four types yields behavioral type combinations and their analysis results. For example, long press-high pressure: marked as "high-difficulty knowledge point," triggering reinforced playback; short press-low pressure: marked as "mastered content," reducing repetitive training.

[0051] In some embodiments, focus analysis involves comparing a calculated focus index with a pre-set focus threshold to obtain analysis results, specifically categorized as high focus and low focus. This includes: based on the focus index distribution of a large number of employees, selecting the top 20% of focus indices as the standard for high focus, and choosing the lowest focus index among these high focus standards as the focus threshold. This needs to be dynamically set according to actual circumstances; for example, a higher focus threshold can be set for training content requiring high focus. The calculated focus index is compared with the pre-set focus threshold to determine if the employee's focus meets the requirements. Training content that does not meet the focus threshold is replayed or subject to specialized training to improve the quality of employee training. Employees are categorized and managed based on the focus analysis results. The training status of each employee is assessed based on the first analysis report, and a subsequent training plan is determined for each employee based on their training status.

[0052] In this embodiment, a first analysis report is obtained based on the baseline operation data of the interactive terminal. The actual training effect of each employee is obtained based on the content of the first analysis report, and the subsequent training content of the employees is adjusted according to the actual training effect.

[0053] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0054] This application utilizes the correspondence between operation area grids and content area grids to accurately record and identify employees' baseline operation data on the interactive terminal, establishes a multi-dimensional behavior model to analyze employee operation data from multiple dimensions, and obtains behavioral intent analysis, focus analysis, and heatmaps to comprehensively display the distribution of employee behavior and focus during the training process; based on the first analysis report, it judges the employee training status, determines subsequent training plans, improves training quality, and achieves in-depth analysis of user behavioral intent and behavior patterns, thereby accurately tracking and optimizing the training effect.

[0055] Example 2: In Example 1, the first analysis report was obtained by acquiring baseline operation data, and employee training was determined and adjusted. However, simple analysis and adjustment alone cannot capture the deep changes in employees. This example makes further improvements based on the above.

[0056] The method further includes: S300: determining the second training content based on the first analysis report and multidimensional analysis rules, and sending the second training content to the training demonstration terminal. Steps S100 and S200 are executed based on the second training content to obtain the second analysis report, which is consistent with the method for obtaining the first analysis report; therefore, this application will not elaborate further on this.

[0057] In some embodiments, the multidimensional profiling rules include: establishing a content region grid-operation behavior matrix, identifying inefficient region grids and obtaining hotspots and blind spots, and establishing a three-dimensional knowledge network based on the hotspots and blind spots and the corresponding behavior type combinations to generate second training content; specifically: establishing a content region grid-operation behavior matrix, generating a mapping table between content region grids and behavior type combinations; obtaining content region grids with a focus index lower than the focus threshold and marking them as inefficient region grids; dividing each inefficient region grid into several micro-grids, and dividing the micro-grids into hotspots and blind spots based on the three-dimensional data of each micro-grid;

[0058] The three-dimensional data is calculated based on the baseline operation data of each microgrid, including operation duration density, operation frequency, and dwell time.

[0059] In some embodiments, a content area grid-behavior type matrix is ​​established based on the content area grids in the first analysis report and the corresponding baseline operation data. The behavior type refers to the various behaviors of the employee corresponding to the baseline operation data, such as long press-high pressure, short press-low pressure, etc. A bidirectional index is established between each content area grid and the corresponding operation area grid. Based on the time axis alignment algorithm, and through a timestamp matching algorithm accurate to 50 milliseconds, the time axis of the training content display is synchronized with the time axis of the user operation record to ensure that the operation data of each content area grid corresponds accurately. The frequency of occurrence of the four basic behavior types (short press, long press, high pressure, low pressure) is counted for each content area grid, and a quantitative mapping table of content area grid and behavior type combination is generated. For example, content area grids with a high frequency of "long press-high pressure" combination are marked as high-difficulty knowledge points; areas where "short press-low pressure" is dominant are marked as mastered content.

[0060] In some embodiments, content areas with a focus index below the focus threshold are identified and marked as inefficient areas. Preferably, the focus threshold needs to be dynamically adjusted when setting it. For example, if three consecutive content areas are found to have a focus index below the current threshold, the focus threshold is automatically lowered to avoid the threshold being too high. If a high stress value is detected and the focus index exceeds the standard, the focus threshold needs to be raised to dynamically adapt to high-performing employees. The specific adjustment needs to be made according to the actual situation.

[0061] In some embodiments, each inefficient region is divided into several microgrids, for example, into 100×100 microgrids; each microgrid records baseline operation data and three-dimensional data, the three-dimensional data including operation duration density, operation frequency, and dwell time, the operation duration density being the proportion of operation time per unit time; the operation frequency being the statistical value of the number of operations, including clicks, swipes, etc.; and the dwell time being the total operation time.

[0062] In some embodiments, the microgrid is divided into hotspot areas and blind areas based on the 3D data of each microgrid. This includes: pre-setting a dwell time baseline value and marking microgrid areas with dwell time less than the dwell time baseline value as blind areas; calculating the thermal value of each microgrid based on the 3D data, arranging the thermal values ​​in descending order, and calculating the average value of the thermal values; and marking according to the average value and marking rules, wherein the marking rules are as follows: if the average value is not less than a pre-set baseline threshold, the microgrid areas corresponding to the top 60% of thermal values ​​are marked as hotspot areas; the remaining microgrid areas are marked as blind areas; if the average value is less than the pre-set baseline threshold, the microgrid areas corresponding to the top 20% of thermal values ​​are marked as hotspot areas; the remaining microgrid areas are marked as blind areas; wherein the thermal value is the product value of the 3D data, and the 3D data needs to be normalized before calculation to ensure that it is within the same dimension for calculation.

[0063] In some embodiments, a three-dimensional knowledge network is established based on hotspot areas and blind spots and their corresponding behavioral types to generate second training content, including: extracting heat features corresponding to hotspot areas and blind spots respectively; obtaining attribution analysis and adjustment strategies based on a pre-established hotspot area attribution rule base and blind spot attribution rule base; obtaining a three-dimensional knowledge network based on heat features, attribution analysis, and adjustment strategies; and adjusting the first training content based on the three-dimensional knowledge network to obtain the second training content.

[0064] In some embodiments, the three-dimensional knowledge network is as follows:

[0065] Thermal characteristic 1: high frequency + low pressure + short dwell time; Attribution analysis 1: icon cognitive ambiguity; Adjustment strategy 1: replace icon design or add floating tooltip;

[0066] Thermal characteristic 2: low frequency + high pressure + long residence; Attribution analysis 2: process step overload; Adjustment strategy 2: split into sub-processes or add flowcharts;

[0067] Based on the corresponding thermal characteristics, attribution analysis, and adjustment strategies, the corresponding training content was reselected, which is the second training content.

[0068] In some embodiments, a hotspot area attribution rule base and a blind area attribution rule base are pre-established. This includes acquiring historical training data, labeling attribution analysis based on the thermal characteristics corresponding to the hotspot area and the blind area respectively, which can be done manually by training experts to label the potential problem type corresponding to each thermal characteristic, i.e., attribution analysis, or automatically labeled by establishing a corresponding analysis model. This application does not impose specific limitations here. Based on the attribution analysis, corresponding adjustment strategies are set, and the association rules between thermal characteristics and attribution analysis are analyzed based on the decision tree algorithm to obtain support. The support is a numerical value used to measure the matching accuracy between thermal characteristics and corresponding attribution analysis. The higher the support, the more accurate the attribution analysis. When using the decision tree algorithm for analysis, it is necessary to analyze based on historical experimental data and actual conditions. The historical experimental data is divided into training datasets and validation datasets. The training dataset is used for algorithm training, and the validation dataset is used for parameter validation. Optimization and adjustment are made based on the validation results. Specific settings need to be made according to actual conditions. This application does not impose specific limitations here. The hotspot area attribution rule base and the blind area attribution rule base are obtained based on the thermal characteristics, attribution analysis, adjustment strategies, and support corresponding to the hotspot area and the blind area respectively. The rule base is a database that includes mapping relationships of features corresponding to different regions. The extracted thermal features are compared one by one with the corresponding rule base to find the corresponding mapping relationship, obtain the attribution analysis, adjustment strategy and corresponding support. If the same thermal features correspond to different attribution analyses, the attribution analysis with the highest support and the corresponding adjustment strategy are selected to generate a three-dimensional knowledge network.

[0069] In some embodiments, the second training content is generated based on the first training content and the first analysis report. Through further analysis of the first analysis report and in combination with the employee's baseline operation data, training content more suitable for the current employee training is obtained, and the second training content is sent to the training display terminal.

[0070] In this embodiment, by integrating click location, duration, and stress value, the system deeply analyzes employees' real training needs, optimizes personalized training paths, and continuously improves training quality through focus index assessment. By selecting and further breaking down content areas into hotspots and blind spots, hotspots accurately pinpoint users' real points of confusion, expose instructional design flaws, and then reduce ineffective repetitive training through targeted reinforcement, improving the accuracy of mastering high-risk operations. Blind spots reveal redundant content or ineffective expression blocks; through content reconstruction, cognitive load is reduced, increasing training efficiency by 20%-40%. By deeply integrating dynamic adjustment data of content areas (grid density, knowledge changes) with operational behavior analysis, the generation of second training content achieves a leap from coarse-grained chapter adaptation to pixel-level interaction optimization. This solves the problem that simply adjusting employee training based on simple analysis cannot capture deep changes in employees. Through more refined analysis and dynamic adjustment of training content, it meets employees' personalized training needs, improving training quality and efficiency.

[0071] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0072] This application deeply analyzes employees' real training needs, generating secondary training content more suitable for current employees; it continuously improves training quality through focus index assessment, accurately identifies employees' real points of confusion through hot spots, and improves the accuracy of mastery; it reveals content redundancy or ineffective expression blocks through blind spots, and reconstructs content to reduce cognitive load and improve training efficiency; it deeply integrates dynamic adjustment data of content area grids with operational behavior analysis, achieving a leap from coarse-grained chapter adaptation to pixel-level interaction optimization.

[0073] Example 3: In the above examples, the second training content was obtained through multi-level analysis, achieving a leap in interactive optimization. This example makes further improvements based on the above content.

[0074] The method further includes: S400: generating a cross-knowledge navigation map and a cross-behavioral pattern curve based on the first and second analysis reports, thereby determining the third training content.

[0075] Specifically, correlation indicators are identified based on the first and second analysis reports to determine the relevant training content. These correlation indicators include regional repetition rate, jump span, and behavioral differences. The regional repetition rate refers to the percentage of times an employee repeatedly operates on the same content area within a unit of time, reflecting persistent deficiencies in knowledge mastery. The jump span refers to the number of content area areas an employee crosses between two adjacent operations, measuring the leaps in knowledge correlation. The behavioral differences refer to the differences in behavioral types (such as long press-high pressure versus short press-low pressure) for different operations within the same content area, resulting in different baseline operational data and reflecting cognitive contradictions.

[0076] The relevance training content is a set of logically related content areas identified from the first and second training content based on relevance indicators. For example, content areas with a repetition rate greater than 30% and a jump span greater than or equal to 2 are collected. Specific screening conditions need to be set according to actual needs. Based on the analysis reports generated by employees in the two training sessions, the corresponding behavior types are analyzed, the corresponding attribution analysis is determined, and the potential problems of employees are identified based on the attribution analysis. Corresponding requirements are set based on the problems. For example, if an employee shows obvious behavioral differences between the two training sessions, the focus should be on the content areas corresponding to the jump span and behavioral differences, and the requirements for the jump span and behavioral differences should be lowered to obtain the content areas.

[0077] The content areas in the relevant training content are broken down into knowledge nodes, and the knowledge nodes are connected based on dependencies. The association weights between different knowledge nodes are calculated, and a cross-knowledge navigation graph is formed based on knowledge nodes, dependencies, and association weights.

[0078] In some embodiments, the knowledge node refers to the smallest knowledge unit in a content area grid, such as the principle of hydraulic valves and alarm code recognition; the dependency relationship refers to the logical dependency relationship between knowledge nodes, including strong dependency and weak dependency. Strong dependency means that a content area grid needs to directly depend on the knowledge content of another content area grid to learn and understand; weak dependency means that a content area grid needs the knowledge content of another content area grid to indirectly supplement it to help learn and understand. The association weight refers to the logical association strength between two content area grids, and is specifically calculated as follows: the cognitive weight and behavioral weight of the employee are preset, and the sum of the cognitive weight and behavioral weight is 1; the cognitive weight is set based on the employee's area repetition rate and jump span. If the area repetition rate and jump span are large, a larger cognitive weight (0.7) is set; the behavioral weight is set based on behavioral differences. If the behavioral differences are large, a larger behavioral weight (0.7) is set.

[0079] By acquiring relevant indicators and related training content, behavioral traversal paths are determined, and a cross-behavioral pattern curve is established. This curve is a three-dimensional curve with time as the horizontal axis and operation duration density, pressure gradient, and focus index as the vertical axis. The pressure gradient refers to the rate of change of pressure values ​​in different content areas. The cross-behavioral pattern curve describes the behavioral evolution of employees in cross-content learning. Specifically, cubic spline interpolation is used to connect the feature values ​​of discrete time windows to generate a smooth curve. The cross-behavioral pattern curve can be briefly divided into three segments for interpretation: Ascending segment: Increased operation duration density combined with a rising pressure gradient indicates entering a deep learning state; Plateau segment: Stable operation duration density combined with small fluctuations in focus indicates a knowledge consolidation period; Descending segment: A sharp drop in pressure gradient indicates distraction or fatigue.

[0080] In some embodiments, determining the third training content includes: identifying the changing trend characteristics of cross-behavioral pattern curves, finding the learning behaviors that best match the changing trend characteristics from cross-knowledge navigation maps, and selecting the corresponding training content as the third training content based on the employee's current knowledge node.

[0081] There is no specific time limit for the unit time described in this application. It needs to be dynamically set according to the actual duration of the training content or the learning time of a single knowledge content. The length of the unit time should be greater than one-tenth of the total training content duration. The specific setting needs to be based on the actual situation. This application does not impose specific restrictions.

[0082] In this embodiment, the structured reorganization of content is achieved through cross-knowledge navigation graphs, and the cognitive evolution law is captured by cross-behavioral pattern curves, ultimately generating dynamically adapted third training content; knowledge association is made explicit, breaking down the content silos of traditional training and constructing a knowledge network that conforms to cognitive laws; at the same time, behavior-driven personalization is achieved, from passively recording operations to actively predicting needs, realizing closed-loop optimization, and automatically adjusting analysis strategies for different business areas to improve the universality of the solution.

[0083] This embodiment addresses the problems of traditional training, such as lack of content relevance, difficulty in capturing the evolution of employee cognition, and inability to dynamically adapt training content to individual employee needs. By using cross-knowledge navigation maps and cross-behavioral pattern curves, it achieves structured reorganization of knowledge, capture of cognitive patterns, and dynamic adaptation of training content, thereby improving training effectiveness and the universality of the program.

[0084] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0085] This application achieves structured reorganization of content through cross-knowledge navigation maps, breaking down content areas in related training content into knowledge nodes and establishing dependencies. It interprets the learning curve segment by segment to understand the employee's learning status, thereby generating dynamically adapted third training content and further improving the accuracy of training content. Based on the changing trend characteristics of cross-behavioral pattern curves and cross-knowledge navigation maps, training content is selected to make knowledge association explicit, achieving the effect of proactive association and dynamic adjustment.

[0086] Example 4: In the above examples, a dynamically adapted third training content is finally generated. This example makes further improvements based on the above content.

[0087] The method further includes: S500: setting personalized trigger conditions based on the third training content, and setting a guided training mechanism according to the personalized trigger conditions; the guided training mechanism automatically compares the changes in employee behavior with the personalized trigger conditions, and automatically adjusts the learning path if the personalized trigger conditions are met.

[0088] The personalized triggering conditions are set based on the slope of the operation duration density, the rate of change of the pressure gradient, and the convergence radius of the focus index, including high-pressure acceleration, dispersion repair, and steady-state transition; the convergence radius of the focus index is the standard deviation of the focus index.

[0089] In some embodiments, the triggering condition for high-pressure acceleration is that the slope of the operation duration density is greater than 0.6 and the rate of change of the pressure gradient is greater than 0.7, in which case advanced content with a cross-knowledge navigation graph association weight greater than 0.8 is pushed; the triggering condition for dispersed repair is that the operation duration density is less than -0.3 and the convergence radius of the focus index is greater than 0.3, in which case pre-knowledge review content is inserted; the triggering condition for steady-state transition is that the convergence radius of the focus index is less than 0.1 and the duration exceeds two units of time, in which case the user jumps to the next knowledge core node.

[0090] In some embodiments, dynamic condition optimization is implemented: a decay mechanism is set, that is, the weight of the triggering condition decays by 10% every month, and its effectiveness needs to be re-verified with new data; a conflict resolution mechanism is set, that is, if multiple conditions are triggered at the same time, the priority needs to be pre-set according to historical experimental data, and the conditions are executed according to the priority (e.g., high pressure acceleration is greater than dispersed repair).

[0091] In this embodiment, by deeply integrating the trend recognition capability across behavioral pattern curves with the structured knowledge association across knowledge navigation maps, a personalized trigger training mechanism that is behavior-driven and knowledge-adaptive is realized. This achieves the effect of moving from static rules to dynamic responses, and automatically evolving with employee skill growth by setting condition thresholds. The deep coupling of guided actions and navigation map paths ensures learning continuity and improves resource allocation efficiency. This solves the problem of traditional training mechanisms lacking personalized triggers and dynamic adjustments. Through personalized trigger conditions and guided training mechanisms, the learning path can be automatically adjusted according to real-time changes in employee behavior, while dynamically optimizing trigger conditions to ensure learning continuity, improve the utilization efficiency of training resources, and better adapt to the needs of employee skill enhancement.

[0092] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0093] This application achieves a personalized trigger training mechanism that combines behavior-driven and knowledge-adaptive approaches by deeply integrating the trend recognition capabilities of cross-behavioral pattern curves with the structured knowledge associations across knowledge navigation maps. It realizes the effect of automatically evolving from static rules to dynamic responses and setting condition thresholds as employees' capabilities grow, further ensuring learning continuity and improving resource allocation efficiency.

[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data processing method for enterprise employee training, applied to a server, a training display terminal, and an interactive terminal, characterized in that, Includes: S100: Obtaining the personnel attribute information of trainees to obtain the corresponding first training content, dividing the first training content into grids to obtain content area grids and content grid sequences, sending the content grid sequence to the interactive terminal; sending the first training content to the training display terminal; S200: Receives baseline operation data from the interactive terminal, obtains combinations of employee behavior types based on a multi-dimensional behavior model, and generates a first analysis report. The first analysis report includes behavior intent analysis, focus analysis, and a heatmap, used to evaluate training effectiveness. The baseline operation data includes operation time, click location, press duration, and pressure value. The behavior intent analysis is obtained by analyzing combinations of multiple behavior types. The focus analysis is obtained by comparing a calculated focus index with a pre-set focus threshold. The heatmap is used to display the distribution of user click locations, pressure values, and pressure duration, and records the focus index at different click locations to identify high-frequency operation areas. S300: Determine the second training content based on the first analysis report and multidimensional analysis rules, and send the second training content to the training demonstration terminal; The multidimensional analysis rules include: establishing a content region grid-operation behavior matrix, identifying inefficient region grids and obtaining hot spots and blind spots, and establishing a three-dimensional knowledge network based on the hot spots, blind spots and corresponding behavior type combinations to generate second training content. Specifically, based on the content region grid-operation behavior matrix, a mapping table between content region grids and behavior type combinations is generated; content region grids with a focus index below the focus threshold are identified and marked as inefficient region grids; each inefficient region grid is divided into several micro-grids, and based on the 3D data of each micro-grid, the micro-grids are divided into hotspot areas and blind areas; a dwell time threshold is preset, and micro-grid areas with dwell time less than the dwell time threshold are marked as blind areas; the heat value of each micro-grid is calculated based on the 3D data, the heat values ​​are arranged in descending order, and the average heat value is calculated. The heat values ​​are labeled using a mean and a labeling rule. The labeling rule is as follows: if the mean value is not less than a preset base threshold, the microgrid regions corresponding to the top 60% of the heat values ​​are labeled as hotspots; the remaining microgrid regions are labeled as blind zones. If the mean value is less than a preset base threshold, the microgrid regions corresponding to the top 20% of the heat values ​​are labeled as hotspots; the remaining microgrid regions are labeled as blind zones. The heat value is the product of three-dimensional data, which is calculated based on the baseline operation data of each microgrid, including operation duration density, operation frequency, and dwell time. S400: Obtain a second analysis report based on the second training content, generate cross-knowledge navigation and cross-behavioral pattern curves based on the first and second analysis reports, and then determine the third training content.

2. The data processing method for enterprise employee training as described in claim 1, characterized in that, The operation time refers to the specific time point when the employee performs the trigger operation; the click location refers to the specific coordinates of the click operation location; the press duration is the duration of a single trigger by the employee; and the pressure value is the pressure applied during a single trigger.

3. The data processing method for enterprise employee training as described in claim 1, characterized in that, The formula for calculating the focus index is as follows: in, It is a focus index; and These are pre-set weighting coefficients. ; It is the pressure value; It is the logarithm of the pressing duration T.

4. The data processing method for enterprise employee training as described in claim 1, characterized in that, A three-dimensional knowledge network is established based on hotspots, blind spots, and corresponding behavioral types to generate secondary training content, including: The thermal features corresponding to hot spots and blind spots are extracted respectively. Attribution analysis and adjustment strategies are obtained based on the pre-established hot spot attribution rule base and blind spot attribution rule base. A three-dimensional knowledge network is obtained based on the thermal features, attribution analysis and adjustment strategies. The first training content is adjusted based on the three-dimensional knowledge network to obtain the second training content. The process involves pre-establishing attribution rule bases for hotspot areas and blind areas. This includes acquiring historical training data, labeling attribution analysis based on the thermal characteristics corresponding to hotspot areas and blind areas respectively, setting corresponding adjustment strategies based on the attribution analysis, and analyzing the association rules between thermal characteristics and attribution analysis using a decision tree algorithm to obtain support. The hotspot area attribution rule base and blind area attribution rule base are then obtained based on the thermal characteristics, attribution analysis, adjustment strategies, and support for hotspot areas and blind areas respectively.

5. The data processing method for enterprise employee training as described in claim 1, characterized in that, Based on the first and second parsing reports, a cross-knowledge navigation map and a cross-behavioral pattern curve are generated, including: Based on the first and second analysis reports, correlation indicators were identified, and relevant training content was determined. The content areas in the relevant training content are broken down into knowledge nodes, and the knowledge nodes are connected based on dependencies. The association weights between different knowledge nodes are calculated, and a cross-knowledge navigation graph is formed based on knowledge nodes, dependencies, and association weights. The behavioral span path is determined based on the correlation indicators and the correlation training content, and a cross-behavioral pattern curve is established. The cross-behavioral pattern curve is a three-dimensional curve with time as the horizontal axis and operation duration density, pressure gradient and focus index as the vertical axis.

6. The data processing method for enterprise employee training as described in claim 5, characterized in that, The third training content is determined by: identifying the changing trend characteristics of cross-behavioral pattern curves, finding the learning behaviors that best match the changing trend characteristics from the cross-knowledge navigation map, and selecting the corresponding training content as the third training content based on the employee's current knowledge nodes.

7. The data processing method for enterprise employee training as described in claim 1, characterized in that, The method further includes: S500: setting personalized trigger conditions based on the third training content, and setting a guided training mechanism according to the personalized trigger conditions; the guided training mechanism automatically compares the changes in employee behavior with the personalized trigger conditions, and automatically adjusts the learning path if the personalized trigger conditions are met.

8. The data processing method for enterprise employee training as described in claim 7, characterized in that, The personalized triggering conditions are set based on the slope of the operation duration density, the rate of change of the pressure gradient, and the convergence radius of the focus index, including high-pressure acceleration, dispersion repair, steady-state transition, and contradiction intervention; the convergence radius of the focus index is the standard deviation of the focus index.

Images