Personnel training data information processing method and system

Abstract

The invention relates to the field of data processing, and provides a personnel training data information processing method and system, and the method comprises the steps: carrying out the semantic embedding coding of the behavior data of a tracked employee object before training and after training through employing a data analysis and processing technology based on deep learning, and automatically judging whether the training effect is positive or negative according to the principal component-based time sequence aggregation representation between the pre-training behavior data semantic embedding characteristics and the post-training behavior data semantic embedding characteristics. Therefore, the behavior pattern change of the employee at multiple time points before and after training can be comprehensively captured instead of only depending on the behavior performance of a single time point, so that the training effect is dynamically and objectively evaluated, and the consistency and fairness of the evaluation process can be ensured by automatically judging the training effect.

Description

Technical Field

[0001] The present application relates to the field of data processing, and more specifically, to a method and system for processing personnel training data information. Background Art

[0002] In modern enterprises, employee training is regarded as a key means to improve employee skills, enhance work efficiency and promote organizational development. As market competition becomes increasingly fierce, enterprises increasingly rely on the professional and innovative capabilities of their employees to maintain their competitive advantage. Effective training can not only help employees master the latest knowledge and technology, but also stimulate their potential and improve their teamwork ability.

[0003] However, traditional training effectiveness evaluation often relies on subjective evaluations by supervisors or colleagues, which may be affected by personal preferences, emotional factors or interpersonal relationships, resulting in evaluation results that are not objective and fair. For example, supervisors may have higher expectations for certain employees and give them higher scores during evaluations. Even if the actual performance of these employees is not outstanding, the relationship between colleagues may affect the fairness of mutual evaluations, resulting in evaluation results that deviate from actual performance. In addition, traditional evaluation methods rely solely on evaluations at a single point in time, making it difficult to capture the long-term impact of training on employee behavior patterns. In other words, evaluations at a single point in time cannot fully reflect changes in employee behavior in different situations, thereby underestimating or overestimating the actual effect of training. This one-sided evaluation method ignores the sustainability and dynamic changes in training effects and cannot provide reliable data support for the long-term development of the company.

[0004] Therefore, an optimized personnel training data information processing solution is needed. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present application provides a method and system for processing personnel training data information.

[0006] According to one aspect of the present application, a method for processing personnel training data information is provided, which includes:

[0007] Obtaining a time queue of pre-training behavior data and a time queue of post-training behavior data of the tracked employee object;

[0008] Semantically embedding the time queue of the pre-training behavior data and the time queue of the post-training behavior data to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data and a time queue of semantic embedding coding vectors of the post-training behavior data;

[0009] Performing behavior data feature temporal aggregation on the time queue of the pre-training behavior data semantic embedding coding vector and the time queue of the post-training behavior data semantic embedding coding vector to obtain the pre-training behavior pattern principal component temporal aggregation representation and the post-training behavior pattern principal component temporal aggregation representation, including: extracting the sequence of the current pre-training behavior data semantic embedding coding vector and the historical pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector, and then performing principal component analysis to obtain the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation; performing principal component temporal aggregation based on gradient attenuation on the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation to obtain the pre-training behavior pattern principal component temporal aggregation representation;

[0010] Performing behavior pattern difference semantic measurement on the principal component temporal aggregation representation of the behavior pattern before training and the principal component temporal aggregation representation of the behavior pattern after training to obtain a semantic representation of the difference between the behavior pattern before and after training;

[0011] Based on the semantic representation of the difference in behavior patterns before and after the training, an evaluation result is obtained.

[0012] According to another aspect of the present application, a personnel training data information processing system is provided, comprising:

[0013] An employee pre-training and post-training behavior data collection module is used to obtain a time queue of pre-training behavior data and a time queue of post-training behavior data of a tracked employee object;

[0014] A semantic embedding coding module for employee behavior data before and after training, used to perform semantic embedding coding on the time queue of the pre-training behavior data and the time queue of the post-training behavior data respectively to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data and a time queue of semantic embedding coding vectors of the post-training behavior data;

[0015] A module for temporal aggregation of behavioral data features before and after employee training, which is used to perform temporal aggregation of behavioral data features on the time queue of the semantic embedding coding vector of the pre-training behavioral data and the time queue of the semantic embedding coding vector of the post-training behavioral data to obtain a temporal aggregation representation of the principal component of the pre-training behavioral pattern and a temporal aggregation representation of the principal component of the post-training behavioral pattern, wherein the module for temporal aggregation of behavioral data features before and after employee training includes: a principal component analysis unit for behavioral data semantic embedding coding vector, which is used to extract the sequence of the current semantic embedding coding vector of the pre-training behavioral data and the sequence of the historical semantic embedding coding vector of the pre-training behavioral data and perform principal component analysis to obtain a sequence of the current semantic principal component representation of the pre-training behavioral data and the sequence of the historical semantic principal component representation of the pre-training behavioral data; a pre-training behavioral pattern principal component temporal aggregation representation generation unit, which is used to perform principal component temporal aggregation based on gradient attenuation on the sequence of the semantic principal component representation of the current semantic principal component representation of the pre-training behavioral data and the sequence of the semantic principal component representation of the historical pre-training behavioral data to obtain the principal component temporal aggregation representation of the pre-training behavioral pattern;

[0016] A module for generating semantic representations of behavioral pattern differences before and after training, used for performing semantic measurement of behavioral pattern differences on the principal component temporal aggregation representation of the behavioral pattern before training and the principal component temporal aggregation representation of the behavioral pattern after training to obtain semantic representations of behavioral pattern differences before and after training;

[0017] The training evaluation effect result generation module is used to obtain the evaluation result based on the semantic representation of the difference in behavior patterns before and after the training.

[0018] This application has significant technical effects due to the adoption of the above technical solutions:

[0019] The personnel training data information processing method and system provided by the present application adopts data analysis and processing technology based on deep learning to perform semantic embedding encoding on the pre-training behavior data and post-training behavior data of the tracked employee object, so as to automatically judge whether the training effect is positive or negative based on the time series aggregation representation between the semantic embedding features of the pre-training behavior data and the semantic embedding features of the post-training behavior data. In this way, it is possible to comprehensively capture the changes in the behavior patterns of employees at multiple time points before and after training, rather than relying solely on the behavioral performance at a single time point, thereby dynamically and objectively evaluating the training effect, and by automatically judging the training effect, the consistency and fairness of the evaluation process can be guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 Flowchart of a method for processing personnel training data information according to an embodiment of the present application.

[0022] Figure 2 Schematic diagram of data flow of a method for processing personnel training data information according to an embodiment of the present application.

[0023] Figure 3 In the personnel training data information processing method according to an embodiment of the present application, a flowchart is provided for performing time series aggregation of behavior data features on the time queue of the semantic embedding coding vector of the pre-training behavior data and the time queue of the semantic embedding coding vector of the post-training behavior data to obtain the time series aggregation representation of the principal components of the pre-training behavior pattern and the time series aggregation representation of the principal components of the post-training behavior pattern.

[0024] Figure 4 In the personnel training data information processing method according to an embodiment of the present application, a flowchart is provided for extracting a sequence of current pre-training behavior data semantic embedding coding vectors and historical pre-training behavior data semantic embedding coding vectors from a time queue of the pre-training behavior data semantic embedding coding vectors, and then performing principal component analysis to obtain a sequence of current pre-training behavior data semantic principal component representations and historical pre-training behavior data semantic principal component representations.

[0025] Figure 5 This is a flowchart of a personnel training data information processing method according to an embodiment of the present application, in which principal component time series aggregation based on gradient decay is performed on the sequence of the semantic principal component representation of the current pre-training behavior data and the semantic principal component representation of the historical pre-training behavior data to obtain the principal component time series aggregation representation of the pre-training behavior pattern.

[0026] Figure 6 4 is a system block diagram of a personnel training data information processing system according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] In today's modern enterprise system, employee training has become a core strategy to improve skills, enhance work efficiency and promote organizational progress. Faced with increasingly fierce market competition, enterprises rely on employees' professional skills and innovative thinking to maintain their market competitiveness. Efficient training can not only enable employees to master the latest knowledge and technology, but also tap their potential and enhance their teamwork ability.

[0029] However, traditional training effectiveness evaluation is usually based on the subjective judgment of superiors or colleagues, which may be affected by personal preferences, emotional factors or interpersonal relationships, thus reducing the objectivity and fairness of the evaluation. For example, superiors may have too high expectations for certain employees, resulting in unrealistically high scores during the evaluation, even if the actual performance of these employees is not outstanding. Mutual evaluation among colleagues may also be affected by interpersonal relationships, making the evaluation results inconsistent with actual performance. In addition, traditional evaluation methods often only focus on performance at a certain point in time, making it difficult to continuously track the long-term impact of training on employee behavior patterns. This single-point evaluation cannot fully capture the behavioral changes of employees in different environments, which may lead to underestimation or overestimation of training effectiveness, ignoring the continuity and dynamism of training effects, and thus failing to provide accurate data support for the long-term development of the enterprise.

[0030] In response to the above technical problems, the technical concept of the present application is to obtain the time queue of the pre-training behavior data and the time queue of the post-training behavior data of the tracked employee object, and use data analysis and processing technology based on deep learning to perform semantic embedding encoding on the pre-training behavior data and the post-training behavior data, so as to automatically judge whether the training effect is positive or negative based on the principal component-based temporal aggregation representation between the semantic embedding features of the pre-training behavior data and the semantic embedding features of the post-training behavior data. In this way, it is possible to comprehensively capture the changes in the behavior patterns of employees at multiple time points before and after training, rather than relying solely on the behavioral performance at a single time point, so as to dynamically and objectively evaluate the training effect. The training effect is automatically judged by conducting behavioral semantic metric learning on employee objects before and after training, thereby ensuring the consistency and fairness of the evaluation process and realizing intelligent personnel training data information processing.

[0031] Figure 1 Flowchart of a method for processing personnel training data information according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the personnel training data information processing method according to the embodiment of the present application. Figure 1 and Figure 2As shown, according to the personnel training data information processing method of the embodiment of the present application, it includes: S110, obtaining the time queue of pre-training behavior data and the time queue of post-training behavior data of the tracked employee object; S120, respectively performing semantic embedding coding on the time queue of pre-training behavior data and the time queue of post-training behavior data to obtain the time queue of pre-training behavior data semantic embedding coding vector and the time queue of post-training behavior data semantic embedding coding vector; S130, performing behavior data feature time series aggregation on the time queue of pre-training behavior data semantic embedding coding vector and the time queue of post-training behavior data semantic embedding coding vector to obtain the time series aggregation representation of the principal component of the pre-training behavior pattern and the time series aggregation representation of the principal component of the post-training behavior pattern; S140, performing behavior pattern difference semantic measurement on the time series aggregation representation of the principal component of the pre-training behavior pattern and the time series aggregation representation of the principal component of the post-training behavior pattern to obtain the semantic representation of the difference between the behavior patterns before and after training; S150, obtaining an evaluation result based on the semantic representation of the difference between the behavior patterns before and after training.

[0032] In step S110, the time queue of the pre-training behavior data and the time queue of the post-training behavior data of the tracked employee object are obtained. It should be understood that the pre-training behavior data refers to various types of work behavior and performance data of the tracked employee object before participating in the training, including but not limited to the number and quality of completed projects, the timeliness of the completion of work tasks, the degree of professional knowledge mastery, etc. The post-training behavior data refers to various types of work and performance data of the tracked employee object after completing the training, and the content included is consistent with the pre-training behavior data. The pre- and post-training behavior data of the employee object is an important basis for evaluating the training effect. In order to reflect the dynamic changes in the behavior of the employee object, provide the continuity of the employee object behavior pattern, and thus more accurately evaluate the training effect, it is necessary to obtain the time queue of the pre-training behavior data and the time queue of the post-training behavior data of the tracked employee object, so as to provide a dynamic perspective for data analysis. By conducting in-depth analysis of these data, the enterprise can evaluate the training effect more objectively and comprehensively, thereby providing a scientific basis for future training decisions.

[0033] In step S120, semantic embedding coding is performed on the time queue of the pre-training behavior data and the time queue of the post-training behavior data, respectively, to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data and a time queue of semantic embedding coding vectors of the post-training behavior data. Specifically, in an embodiment of the present application, semantic embedding coding is performed on the time queue of the pre-training behavior data and the time queue of the post-training behavior data, respectively, to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data and a time queue of semantic embedding coding vectors of the post-training behavior data, including: using a behavior data semantic embedding matrix to perform semantic embedding coding on each pre-training behavior data in the time queue of the pre-training behavior data to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data; using the behavior data semantic embedding matrix to perform semantic embedding coding on each post-training behavior data in the time queue of the post-training behavior data to obtain a time queue of semantic embedding coding vectors of the post-training behavior data.

[0034] It should be understood that, considering that each pre-training behavior data in the time queue of pre-training behavior data is high-dimensional and relatively complex to process, in order to map the pre-training behavior data to a low-dimensional space, in order to reduce the complexity of calculation and thus improve the efficiency of data processing, and at the same time to be able to capture the semantic information between the pre-training behavior data, so that the processed data can better reflect the core characteristics of the original behavior data, and thus better characterize the work behavior performance of employees before training. Based on this, in the technical solution of the present application, the semantic embedding matrix of the behavior data is used to perform semantic embedding encoding on each pre-training behavior data in the time queue of the pre-training behavior data to obtain a time queue of the semantic embedding encoding vector of the pre-training behavior data. That is, the semantic embedding matrix of the behavior data converts the original pre-training behavior data into a vector representation, and can extract the pre-training employee behavior performance characteristics implicit in the data. Through semantic embedding encoding, the pre-training behavior of employees can be captured more accurately, thereby providing more accurate and comprehensive support for the final training effect evaluation. Similarly, the semantic embedding matrix of the behavior data is used to perform semantic embedding encoding on each post-training behavior data in the time queue of the post-training behavior data to obtain a time queue of the semantic embedding encoding vector of the post-training behavior data. By semantically embedding and encoding the post-training behavior data in this way, we can capture the semantic relationship between the post-training behavior data, thereby more accurately reflecting the actual work behavior performance of employees after training.

[0035] In step S130, the time queue of the semantic embedding coding vector of the pre-training behavior data and the time queue of the semantic embedding coding vector of the post-training behavior data are subjected to time series aggregation of behavior data features to obtain a time series aggregation representation of the principal components of the pre-training behavior pattern and a time series aggregation representation of the principal components of the post-training behavior pattern. Accordingly, considering that the vectors in each time queue of the time queue of the semantic embedding coding vector of the pre-training behavior data and the time queue of the semantic embedding coding vector of the post-training behavior data represent the semantic information of the behavior data at a time point. Moreover, there is an inherent time dependency in the semantic information of the behavior data at different time points, that is, the behavior at a certain moment may be affected by the behavior at the previous moment. Therefore, in order to capture and quantify the changes in the behavior pattern over time, so as to more accurately evaluate the effect of the training, in the technical solution of the present application, the time queue of the semantic embedding coding vector of the pre-training behavior data and the time queue of the semantic embedding coding vector of the post-training behavior data are subjected to time series aggregation of behavior data features to obtain a time series aggregation representation of the principal components of the pre-training behavior pattern and a time series aggregation representation of the principal components of the post-training behavior pattern.

[0036] In particular, the time queue of the semantic embedding coding vector of the pre-training behavior data is taken as an example for specific explanation. Specifically, first, the current pre-training behavior data semantic embedding coding vector (current vector) and the sequence of historical pre-training behavior data semantic embedding coding vectors (sequence of historical vectors) are extracted from the time queue of the semantic embedding coding vector of the pre-training behavior data and principal component analysis is performed. It should be understood that principal component analysis (PCA) is a dimensionality reduction technique that can extract the most representative features (principal components) from high-dimensional data. That is, by performing PCA on the sequence of current vectors and historical vectors, those features that are most important to the pre-training behavior pattern can be identified, thereby obtaining the current pre-training behavior data semantic principal component representation vector as the current pre-training behavior data semantic principal component representation and the sequence of historical pre-training behavior data semantic principal component representation vectors as the sequence of historical pre-training behavior data semantic principal component representation.

[0037] Next, the sequence of the semantic principal component representation of the current pre-training behavior data and the semantic principal component representation of the historical pre-training behavior data is subjected to principal component temporal aggregation based on gradient attenuation to obtain a semantic representation vector of the difference between the behavior patterns before and after training as the semantic representation of the difference between the behavior patterns before and after training. Specifically, in order to further quantify and understand the degree of change of the behavior pattern over time and its influencing factors, the absolute factor of the temporal propagation attenuation entropy of each historical vector relative to the current vector is calculated to obtain a sequence of absolute factors of the temporal propagation attenuation entropy of the pre-training behavior data. That is, the absolute factor of the temporal propagation attenuation entropy can evaluate the influence of the pre-training behavior characteristics at different time points on the current pre-training behavior state, so as to better understand which factors play a key role in the change of pre-training behavior. Afterwards, in order to further enhance the ability to transmit and understand the semantics of the pre-training behavior data over a long span, each absolute factor of the temporal propagation attenuation entropy is adjusted by introducing the time span between each historical vector and the current vector. This adjustment allows the model to better handle the relationship between long and short-term memory, ensuring that even if the time interval between each node vector is long, the significant difference between them is regarded as a reasonable fluctuation to obtain a sequence of the temporal span modulation propagation attenuation entropy factor of the pre-training behavior data. Then, the information transmission screening module based on the gating function is used to adaptively select and assign corresponding weights to each modulated factor, that is, dynamically adjust the information transmission according to the importance of the current information to filter out truly meaningful information and suppress irrelevant or redundant information, so as to obtain a sequence of the modulation propagation attenuation weights of the temporal span of the pre-training behavior data. Next, each weight in the weight sequence is used as the weight value to perform weighted aggregation on the sequence of historical vectors to more accurately reflect the influence of the historical pre-training behavior data on the current pre-training behavior data, thereby generating a significant transfer aggregation representation vector of the semantic principal component of the historical pre-training behavior data. Finally, the historical aggregation representation vector and the current pre-training behavior data semantic embedding encoding vector are added by position to obtain the final pre-training behavior pattern principal component temporal aggregation representation vector. In this way, the continuity and change trend of the pre-training behavior pattern are better reflected, which helps to improve the model's ability to recognize changes in the pre-training behavior pattern.

[0038] Specifically, Figure 3 The flowchart of the personnel training data information processing method according to the embodiment of the present application is to perform behavior data feature time series aggregation on the time queue of the semantic embedding coding vector of the pre-training behavior data and the time queue of the semantic embedding coding vector of the post-training behavior data to obtain the time series aggregation representation of the principal component of the pre-training behavior pattern and the time series aggregation representation of the principal component of the post-training behavior pattern. Figure 3As shown, the time queue of the pre-training behavior data semantic embedding coding vector and the time queue of the post-training behavior data semantic embedding coding vector are subjected to time series aggregation of behavior data features to obtain a time series aggregation representation of the pre-training behavior pattern principal component and a time series aggregation representation of the post-training behavior pattern principal component, including: S131, extracting the sequence of the current pre-training behavior data semantic embedding coding vector and the sequence of the historical pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector, and then performing principal component analysis to obtain a sequence of the current pre-training behavior data semantic principal component representation and the sequence of the historical pre-training behavior data semantic principal component representation; S132, performing principal component time series aggregation based on gradient attenuation on the sequence of the current pre-training behavior data semantic principal component representation and the sequence of the historical pre-training behavior data semantic principal component representation to obtain the time series aggregation representation of the pre-training behavior pattern principal component.

[0039] More specifically, Figure 4 The present invention is a flowchart of a method for processing personnel training data information according to an embodiment of the present application, wherein a sequence of current pre-training behavior data semantic embedding coding vectors and historical pre-training behavior data semantic embedding coding vectors is extracted from a time queue of the pre-training behavior data semantic embedding coding vectors, and then a principal component analysis is performed to obtain a sequence of current pre-training behavior data semantic principal component representations and historical pre-training behavior data semantic principal component representations. Figure 4 As shown, extracting the sequence of the current pre-training behavior data semantic embedding coding vector and the historical pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector and then performing principal component analysis to obtain the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation, including: S1311, extracting the current pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector, and performing eigenvalue-based principal component extraction on the current pre-training behavior data semantic embedding coding vector to obtain the current pre-training behavior data semantic principal component representation vector as the semantic principal component representation of the current pre-training behavior data; S1312, define other pre-training behavior data semantic embedding coding vectors in the time queue of the pre-training behavior data semantic embedding coding vector as historical pre-training behavior data semantic embedding coding vectors to obtain a sequence of the historical pre-training behavior data semantic embedding coding vectors, and perform eigenvalue-based principal component extraction on each historical pre-training behavior data semantic embedding coding vector in the sequence of historical pre-training behavior data semantic embedding coding vectors to obtain a sequence of historical pre-training behavior data semantic principal component representation vectors as the sequence of historical pre-training behavior data semantic principal component representations.

[0040] More specifically, Figure 5The present invention is a flowchart of performing principal component temporal aggregation based on gradient decay on the sequence of the semantic principal component representation of the current pre-training behavior data and the semantic principal component representation of the historical pre-training behavior data in the personnel training data information processing method according to the embodiment of the present application to obtain the principal component temporal aggregation representation of the pre-training behavior pattern. Figure 5 As shown, the principal component temporal aggregation based on gradient attenuation is performed on the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation to obtain the principal component temporal aggregation representation of the pre-training behavior pattern, including: S1321, calculating the absolute factor of the temporal propagation attenuation entropy of each historical pre-training behavior data semantic principal component representation vector in the sequence of the historical pre-training behavior data semantic principal component representation vector relative to the current pre-training behavior data semantic principal component representation vector to obtain a sequence of absolute factors of the temporal propagation attenuation entropy of the pre-training behavior data; S1322, calculating the absolute factor of the temporal propagation attenuation entropy of each historical pre-training behavior data semantic principal component representation vector in the sequence of the historical pre-training behavior data semantic principal component representation vector based on the time span between each historical pre-training behavior data semantic principal component representation vector in the sequence of the historical pre-training behavior data semantic principal component representation vector and the current pre-training behavior data semantic principal component representation vector. The absolute factor of the temporal propagation attenuation entropy is modulated in the time dimension to obtain a sequence of the temporal span modulated propagation attenuation entropy factors of the pre-training behavior data; S1323, the sequence of the temporal span modulated propagation attenuation entropy factors of the pre-training behavior data is input into the information transfer screening module based on the gating function to obtain a sequence of the temporal span modulated propagation attenuation weights of the pre-training behavior data; S1324, based on the sequence of the temporal span modulated propagation attenuation weights of the pre-training behavior data, the weighted sum of the sequence of the semantic principal component representation vectors of the historical pre-training behavior data is calculated to obtain the significant transfer aggregation representation vector of the semantic principal component of the historical pre-training behavior data; S1325, the positional sum of the significant transfer aggregation representation vector of the semantic principal component of the historical pre-training behavior data and the semantic principal component representation vector of the current pre-training behavior data is calculated to obtain the temporal aggregation representation vector of the principal component of the pre-training behavior pattern as the temporal aggregation representation of the principal component of the pre-training behavior pattern.

[0041] More specifically, in an embodiment of the present application, the absolute factors of the temporal propagation attenuation entropy of each historical pre-training behavior data semantic principal component representation vector in the sequence of the historical pre-training behavior data semantic principal component representation vectors relative to the current pre-training behavior data semantic principal component representation vector are calculated to obtain a sequence of absolute factors of the temporal propagation attenuation entropy of the pre-training behavior data, including: calculating the positional division of the historical pre-training behavior data semantic principal component representation vector and the current pre-training behavior data semantic principal component representation vector to obtain the pre-training behavior data semantic principal component attenuation vector; calculating the base-two logarithmic function value of the absolute value of each eigenvalue of the pre-training behavior data semantic principal component attenuation vector to obtain the pre-training behavior data semantic principal component attenuation logarithmic vector; calculating the positional dot product of the historical pre-training behavior data semantic principal component representation vector and the pre-training behavior data semantic principal component attenuation logarithmic vector, and performing positional multiplication on the obtained dot product vector to obtain the pre-training behavior data temporal propagation attenuation value; calculating an exponential function with the natural constant e as the base, the pre-training behavior data temporal propagation attenuation value being the exponent to obtain the pre-training behavior data temporal propagation attenuation entropy absolute factors.

[0042] More specifically, in an embodiment of the present application, based on the time span between each historical pre-training behavior data semantic principal component representation vector in the sequence of historical pre-training behavior data semantic principal component representation vectors and the current pre-training behavior data semantic principal component representation vector, each pre-training behavior data temporal propagation attenuation entropy absolute factor in the sequence of pre-training behavior data temporal propagation attenuation entropy absolute factors is modulated in the time dimension to obtain a sequence of pre-training behavior data temporal span modulation propagation attenuation entropy factors, including: subtracting the timestamp of the current pre-training behavior data semantic principal component representation vector from the timestamp of the historical pre-training behavior data semantic principal component representation vector and rounding down to obtain the pre-training behavior data time span value; calculating an exponential function with the natural constant e as the base and the pre-training behavior data time span value as the exponent to obtain the pre-training behavior data time span modulation value; dividing the pre-training behavior data temporal propagation attenuation entropy absolute factor corresponding to the historical pre-training behavior data semantic principal component representation vector by the pre-training behavior data time span modulation value to obtain the pre-training behavior data temporal span modulation propagation attenuation entropy factor.

[0043] More specifically, in an embodiment of the present application, the sequence of the timing span modulation propagation attenuation entropy factors of the pre-training behavior data is input into an information transfer screening module based on a gating function to obtain a sequence of the timing span modulation propagation attenuation weights of the pre-training behavior data, including: comparing each of the timing span modulation propagation attenuation entropy factors of the pre-training behavior data in the sequence of the timing span modulation propagation attenuation entropy factors of the pre-training behavior data with a predetermined threshold to obtain the sequence of the timing span modulation propagation attenuation weights of the pre-training behavior data; wherein, in response to the timing span modulation propagation attenuation entropy factor of the pre-training behavior data being greater than the predetermined threshold, the timing span modulation propagation attenuation entropy factor of the pre-training behavior data being greater than the predetermined threshold is input into a sigmoid function; in response to the timing span modulation propagation attenuation entropy factor of the pre-training behavior data being less than or equal to the predetermined threshold, the timing span modulation propagation attenuation entropy factor of the pre-training behavior data being less than or equal to the predetermined threshold is set to zero.

[0044] In the embodiment of the present application, specifically, the time series aggregation of the behavior data features is performed on the time queue of the semantic embedding coding vector of the pre-training behavior data, which can be expressed by the formula:

[0045] X={x1,x2,...,xi,...,x p}

[0046] v i = PCA(x i )

[0047]

[0048] V = {v1, v2, ..., v i , ..., v p}

[0049]

[0050]

[0051]

[0052]

[0053] Where X is the time sequence of the semantic embedding encoding vector of the pre-training behavior data, x1, x2, ..., x i , ..., x p are the 1st, 2nd, ..., ith, ..., pth pre-training behavior data semantic embedding encoding vectors in the time queue of the pre-training behavior data semantic embedding encoding vectors, PCA(x i ) is x iPerform principal component extraction based on eigenvalues, v i For x i The corresponding semantic principal component representation vector of pre-training behavior data, x i T For x i The transposed vector of n is the number of feature values ​​in the semantic embedding encoding vector of the behavioral data before training, C i is x i The corresponding pre-training behavior data semantic covariance matrix, U i is x i The corresponding pre-training behavior data semantic principal component orthogonal matrix, U i T For U i The transposed matrix, v i1 , v i2 , ..., v im is the semantic principal component vector of each pre-training behavior data in the sequence of semantic principal component vectors of pre-training behavior data, Λ i For x i The corresponding pre-training behavior data semantic diagonal matrix, diag(λ i1 ,λ i2 , ..., λ imm ) is the matrix whose diagonal elements are λ i1 ,λ i2 , ..., λ im The pre-training behavior data semantic diagonal matrix, λ i1 ,λ i2 , ..., λ im are the weight values ​​of the semantic principal component vectors of each pre-training behavior data, argmax k (·) is the k value corresponding to the maximum value returned, j is the maximum approximate matching value, V is the sequence of semantic principal component representation vectors of the pre-training behavior data, v1, v2, ..., v i , ..., v p are respectively the 1st, 2nd, ..., ith, ..., pth semantic principal component representation vectors of the pre-training behavior data in the sequence of the semantic principal component representation vectors of the pre-training behavior data, v i The eigenvalues ​​at each position in , v p The eigenvalues ​​of each position in, L is v i The number of eigenvalues ​​in the equation, log represents the logarithmic function value with base 2, exp(·) represents the exponential function with the natural constant e as the base, d i v i and v p The absolute factor of decay entropy between the pre-training behavioral data temporal propagation, tp and t i denote the timestamps of the semantic principal component representation vectors of the p-th and i-th pre-training behavioral data, respectively. For floor rounding operation, e (i→p) Yes i and v p The pre-training behavioral data temporal span modulation propagation attenuation entropy factor between them, mask(·) is the masking process, sigmoid(·) is the sigmoid function, θ is the predetermined threshold, w i is the temporal span modulation propagation attenuation weight of each pre-training behavior data in the sequence of the temporal span modulation propagation attenuation weight of the pre-training behavior data, p is the number of vectors in the sequence of the semantic principal component representation vector of the pre-training behavior data, v h is the principal component temporal aggregation representation vector of the pre-training behavior pattern. Similarly, the processing method of the time queue of the semantic embedding encoding vector of the post-training behavior data is also as shown in the above formula.

[0054] In step S140, the behavior pattern difference semantic measurement is performed on the main component temporal aggregation representation of the behavior pattern before training and the main component temporal aggregation representation of the behavior pattern after training to obtain the semantic representation of the difference between the behavior pattern before and after training. Specifically, in an embodiment of the present application, the behavior pattern difference semantic measurement is performed on the main component temporal aggregation representation of the behavior pattern before training and the main component temporal aggregation representation of the behavior pattern after training to obtain the semantic representation of the difference between the behavior pattern before and after training, including: inputting the main component temporal aggregation representation vector of the behavior pattern before training and the main component temporal aggregation representation vector of the behavior pattern after training into the behavior pattern difference semantic measurement network to obtain the semantic representation vector of the difference between the behavior pattern before and after training as the semantic representation of the difference between the behavior pattern before and after training. It should be understood that in order to accurately measure the degree of change of the behavior pattern before and after training, so as to provide a quantitative indicator to reflect whether the training has brought about substantial changes, in the technical solution of the present application, the main component temporal aggregation representation vector of the behavior pattern before training and the main component temporal aggregation representation vector of the behavior pattern after training are input into the behavior pattern difference semantic measurement network to obtain the semantic representation vector of the difference between the behavior pattern before and after training. In particular, in a specific embodiment of the present application, the difference semantic representation vector of the behavior pattern before and after training can be obtained by calculating the positional difference between the principal component temporal aggregation representation vector of the behavior pattern before training and the principal component temporal aggregation representation vector of the behavior pattern after training.

[0055] In step S150, the behavior pattern difference semantic measurement is performed on the time series aggregation representation of the principal component of the behavior pattern before training and the time series aggregation representation of the principal component of the behavior pattern after training to obtain the semantic representation of the difference between the behavior patterns before and after training. Specifically, in an embodiment of the present application, based on the semantic representation of the difference between the behavior patterns before and after training, an evaluation result is obtained, including: inputting the semantic representation vector of the difference between the behavior patterns before and after training into a training effect evaluator based on a classifier to obtain the evaluation result, and the evaluation result is used to indicate whether the training effect is positive or negative. That is, the semantic representation of the difference between the behavior patterns before and after training obtained by semantic measurement using the time series aggregation representation vector of the principal component of the behavior pattern before training and the time series aggregation representation vector of the principal component of the behavior pattern after training is classified, so as to automatically determine whether the training effect is positive or negative. In this way, the changes in the behavior patterns of employees at multiple time points before and after training can be fully captured, rather than relying solely on the behavioral performance at a single time point, thereby dynamically and objectively evaluating the training effect. The training effect is automatically judged by learning behavioral semantic metrics of employees before and after training, thereby ensuring the consistency and fairness of the evaluation process and realizing intelligent personnel training data information processing.

[0056] In particular, the present application takes into account that the time queue of the pre-training behavior data semantic embedding coding vector and the time queue of the post-training behavior data semantic embedding coding vector respectively represent the time queue of the pre-training behavior semantic embedding coding features of the tracked employee object and the time queue of the post-training behavior semantic coding features of the tracked employee object. When the behavior data features are time-series aggregated, the differences in feature principal component extraction and gradient attenuation will cause the pre-training behavior pattern principal component time-series aggregation representation vector and the post-training behavior pattern principal component time-series aggregation representation vector to have more significant fine-grained structural differences in the distribution of feature manifolds in the high-dimensional feature space. In this way, the feature manifold of the pre- and post-training behavior pattern difference semantic representation vector obtained by inputting the pre-training behavior pattern principal component time-series aggregation representation vector and the post-training behavior pattern principal component time-series aggregation representation vector into the behavior pattern difference semantic measurement network will also have a complex aggregation space structure, thereby expecting to improve its classification regression convergence and generalization effects under the complex aggregation space structure.

[0057] Preferably, in an example of the present application, the evaluation result of the semantic representation vector of the difference in behavior patterns before and after training is obtained by a classifier-based training effect evaluator, including: calculating the sum of the absolute values ​​of each eigenvalue of the semantic representation vector of the difference in behavior patterns before and after training to obtain a first semantic representation space structure value of the difference in behavior patterns before and after training, and calculating the square root of the sum of the squares of each eigenvalue of the semantic representation vector of the difference in behavior patterns before and after training to obtain a second semantic representation space structure value of the difference in behavior patterns before and after training; multiplying each eigenvalue of the semantic representation vector of the difference in behavior patterns before and after training with the first semantic representation space structure value of the difference in behavior patterns before and after training and the second semantic representation space structure value of the difference in behavior patterns before and after training to obtain a first semantic representation structure reference value of the difference in behavior patterns before and after training and a second semantic representation structure reference value of the difference in behavior patterns before and after training corresponding to each eigenvalue; multiplying each eigenvalue of the semantic representation vector of the difference in behavior patterns before and after training with the length of the semantic representation vector of the difference in behavior patterns before and after training and the square root of the length to obtain a value corresponding to each eigenvalue. The corresponding first semantic representation scale transformation value of the difference in behavior pattern before and after training and the second semantic representation scale transformation value of the difference in behavior pattern before and after training; dividing the first semantic representation structure reference value of the difference in behavior pattern before and after training by the difference between the first semantic representation space structure value of the difference in behavior pattern before and after training and the first semantic representation scale transformation value of the difference in behavior pattern before and after training to obtain the first semantic representation transformation adjustment value of the difference in behavior pattern before and after training; dividing the second semantic representation structure reference value of the difference in behavior pattern before and after training by the difference between the second semantic representation space structure value of the difference in behavior pattern before and after training and the second semantic representation scale transformation value of the difference in behavior pattern before and after training to obtain the second semantic representation transformation adjustment value of the difference in behavior pattern before and after training; calculating the weighted sum of the first semantic representation transformation adjustment value of the difference in behavior pattern before and after training and the second semantic representation transformation adjustment value of the difference in behavior pattern before and after training to obtain each eigenvalue of the optimized semantic representation vector of the difference in behavior pattern before and after training; passing the optimized semantic representation vector of the difference in behavior pattern before and after training through a classifier-based training effect evaluator to obtain an evaluation result.

[0058] Here, the optimization representation of the semantic representation vector of the difference in behavior patterns before and after training is:

[0059]

[0060]

[0061] v 1i =(α×v i ) / (α-L×v i )

[0062]

[0063]

[0064]

[0065]

[0066]

[0067] Wherein, V is the semantic representation vector of the difference between the behavioral patterns before and after training, L is the length of the semantic representation vector of the difference between the behavioral patterns before and after training, and v i is the eigenvalue of the ith position in the semantic representation vector of the difference between the behavioral patterns before and after training, α is the spatial structure value of the semantic representation of the first difference between the behavioral patterns before and after training, β is the spatial structure value of the semantic representation of the second difference between the behavioral patterns before and after training, and v 1i is the first pre-training behavior pattern difference semantic representation transformation adjustment value of the ith position in the first pre-training behavior pattern difference semantic representation transformation adjustment vector, v 2i is the second before-and-after training behavior pattern difference semantic representation transformation adjustment value of the ith position in the second before-and-after training behavior pattern difference semantic representation transformation adjustment vector, V1 is the first before-and-after training behavior pattern difference semantic representation transformation adjustment vector, V2 is the second before-and-after training behavior pattern difference semantic representation transformation adjustment vector, ω represents the weighted hyperparameter, represents point-wise addition, ⊙ represents point-wise multiplication, and V′ is the optimized semantic representation vector of the behavioral pattern difference before and after training.

[0068] That is, in the preferred example, for the spatial structure information of the feature set of the semantic representation vector of the difference in behavior pattern before and after training in the high-dimensional space, the scale-based frame transformation of each feature value of the semantic representation vector of the difference in behavior pattern before and after training is performed by taking the class norm space structured representation of the semantic representation vector of the difference in behavior pattern before and after training as the reference window, and realizing the frame attention weight adjustment based on the spatial structure of each feature value of the semantic representation vector of the difference in behavior pattern before and after training, so as to ensure the invariance of the spatial transformation (translation, scaling and rotation) of the semantic representation vector of the difference in behavior pattern before and after training under the interaction of feature space, thereby improving the convergence and generalization effect of the classification regression of the feature set of the semantic representation vector of the difference in behavior pattern before and after training under the complex spatial structure representation, and improving the accuracy of the evaluation result obtained by the training effect evaluator based on the classifier. In this way, the changes in the behavior patterns of employees at multiple time points before and after training can be fully captured, rather than relying solely on the behavioral performance at a single time point, so as to dynamically and objectively evaluate the training effect. The training effect is automatically judged by learning the behavioral semantic metric of employee objects based on before and after training, thereby ensuring the consistency and fairness of the evaluation process and realizing intelligent personnel training data information processing.

[0069] In summary, the personnel training data information processing method based on the embodiment of the present application is explained, which uses data analysis and processing technology based on deep learning to semantically embed the pre-training behavior data and post-training behavior data of the tracked employee object, so as to automatically judge whether the training effect is positive or negative based on the time series aggregation representation based on the principal component between the semantic embedding features of the pre-training behavior data and the semantic embedding features of the post-training behavior data. In this way, it is possible to comprehensively capture the changes in the behavior patterns of employees at multiple time points before and after training, rather than relying solely on the behavioral performance at a single time point, thereby dynamically and objectively evaluating the training effect, and by automatically judging the training effect, the consistency and fairness of the evaluation process can be guaranteed.

[0070] Figure 6 FIG. 1 is a system block diagram of a personnel training data information processing system according to an embodiment of the present application. Figure 6As shown, according to the embodiment of the present application, the personnel training data information processing system 100 includes: an employee pre-training and post-training behavior data acquisition module 110, which is used to obtain the time queue of pre-training behavior data and the time queue of post-training behavior data of the tracked employee object; an employee pre-training and post-training behavior data semantic embedding coding module 120, which is used to respectively perform semantic embedding coding on the time queue of the pre-training behavior data and the time queue of the post-training behavior data to obtain the time queue of the pre-training behavior data semantic embedding coding vector and the time queue of the post-training behavior data semantic embedding coding vector; an employee pre-training and post-training behavior data feature time series aggregation module 130, which is used to perform behavior data feature time series aggregation on the time queue of the pre-training behavior data semantic embedding coding vector and the time queue of the post-training behavior data semantic embedding coding vector to obtain the pre-training behavior pattern principal component time series aggregation representation and the post-training behavior pattern principal component time series aggregation representation; a pre-training and post-training behavior pattern difference semantic representation generation module 140, which is used to perform the pre-training behavior pattern principal component time series aggregation The combined representation and the time series aggregation representation of the principal component of the post-training behavior pattern are used to perform semantic measurement of the behavior pattern difference to obtain the semantic representation of the behavior pattern difference before and after training; a training evaluation effect result generation module 150 is used to obtain an evaluation result based on the semantic representation of the behavior pattern difference before and after training; wherein the time series aggregation module of the employee behavior data features before and after training includes: a behavior data semantic embedding coding vector principal component analysis unit, which is used to extract the sequence of the current pre-training behavior data semantic embedding coding vector and the historical pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector and then perform principal component analysis to obtain the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation; a pre-training behavior pattern principal component time series aggregation representation generation unit, which is used to perform principal component time series aggregation based on gradient attenuation on the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation to obtain the pre-training behavior pattern principal component time series aggregation representation.

[0071] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the personnel training data information processing system 100 have been described in detail above. Figures 1 to 5 The description of the personnel training data information processing method has been described in detail, and therefore, its repeated description will be omitted.

[0072] In summary, the personnel training data information processing system 100 based on the embodiment of the present application is explained, which uses data analysis and processing technology based on deep learning to perform semantic embedding encoding on the pre-training behavior data and post-training behavior data of the tracked employee object, so as to automatically judge whether the training effect is positive or negative based on the time series aggregation representation based on the principal component between the semantic embedding features of the pre-training behavior data and the semantic embedding features of the post-training behavior data. In this way, it is possible to fully capture the changes in the behavior patterns of employees at multiple time points before and after training, rather than relying solely on the behavioral performance at a single time point, thereby dynamically and objectively evaluating the training effect, and by automatically judging the training effect, the consistency and fairness of the evaluation process can be guaranteed.

[0073] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. A personnel training data information processing method, characterized in that: include: Obtaining a time queue of pre-training behavior data and a time queue of post-training behavior data of the tracked employee object; Semantically embedding the time queue of the pre-training behavior data and the time queue of the post-training behavior data to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data and a time queue of semantic embedding coding vectors of the post-training behavior data; Performing behavior data feature temporal aggregation on the time queue of the pre-training behavior data semantic embedding coding vector and the time queue of the post-training behavior data semantic embedding coding vector to obtain the pre-training behavior pattern principal component temporal aggregation representation and the post-training behavior pattern principal component temporal aggregation representation, including: extracting the sequence of the current pre-training behavior data semantic embedding coding vector and the historical pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector, and then performing principal component analysis to obtain the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation; performing principal component temporal aggregation based on gradient attenuation on the sequence of the current pre-training behavior data semantic principal component representation and the historical pre-training behavior data semantic principal component representation to obtain the pre-training behavior pattern principal component temporal aggregation representation; Performing behavior pattern difference semantic measurement on the principal component temporal aggregation representation of the behavior pattern before training and the principal component temporal aggregation representation of the behavior pattern after training to obtain a semantic representation of the difference between the behavior pattern before and after training; Based on the semantic representation of the difference in behavior patterns before and after the training, an evaluation result is obtained.

2. The personnel training data information processing method according to claim 1, characterized in that: Semantically embedding the time queue of the pre-training behavior data and the time queue of the post-training behavior data to obtain a time queue of the pre-training behavior data semantic embedding encoding vector and a time queue of the post-training behavior data semantic embedding encoding vector, respectively, including: Using a behavior data semantic embedding matrix, semantic embedding coding is performed on each pre-training behavior data in the time queue of the pre-training behavior data to obtain a time queue of the pre-training behavior data semantic embedding coding vector; The behavior data semantic embedding matrix is ​​used to perform semantic embedding encoding on each post-trained behavior data in the time queue of the post-trained behavior data to obtain a time queue of the post-trained behavior data semantic embedding encoding vectors.

3. The personnel training data information processing method according to claim 2 is characterized in that: The method further comprises: extracting a sequence of a current pre-training behavior data semantic embedding coding vector and a sequence of a historical pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector, and then performing principal component analysis to obtain a sequence of a current pre-training behavior data semantic principal component representation and a sequence of a historical pre-training behavior data semantic principal component representation, including: Extracting the current pre-training behavior data semantic embedding coding vector from the time queue of the pre-training behavior data semantic embedding coding vector, and performing eigenvalue-based principal component extraction on the current pre-training behavior data semantic embedding coding vector to obtain the current pre-training behavior data semantic principal component representation vector as the current pre-training behavior data semantic principal component representation; Define other pre-training behavior data semantic embedding coding vectors in the time queue of the pre-training behavior data semantic embedding coding vectors as historical pre-training behavior data semantic embedding coding vectors to obtain a sequence of historical pre-training behavior data semantic embedding coding vectors, and perform eigenvalue-based principal component extraction on each historical pre-training behavior data semantic embedding coding vector in the sequence of historical pre-training behavior data semantic embedding coding vectors to obtain a sequence of historical pre-training behavior data semantic principal component representation vectors as the sequence of historical pre-training behavior data semantic principal component representations.

4. The personnel training data information processing method according to claim 3 is characterized in that: The method of performing principal component temporal aggregation based on gradient decay on the sequence of the semantic principal component representation of the current pre-training behavior data and the semantic principal component representation of the historical pre-training behavior data to obtain the principal component temporal aggregation representation of the pre-training behavior pattern includes: Calculating the absolute factors of temporal propagation decay entropy of each historical pre-training behavior data semantic principal component representation vector in the sequence of historical pre-training behavior data semantic principal component representation vectors relative to the current pre-training behavior data semantic principal component representation vector to obtain a sequence of absolute factors of temporal propagation decay entropy of pre-training behavior data; Based on the time span between each historical pre-training behavior data semantic principal component representation vector in the sequence of historical pre-training behavior data semantic principal component representation vectors and the current pre-training behavior data semantic principal component representation vector, each pre-training behavior data temporal propagation attenuation entropy absolute factor in the sequence of pre-training behavior data temporal propagation attenuation entropy absolute factors is modulated in time dimension to obtain a sequence of pre-training behavior data temporal span modulated propagation attenuation entropy factors; Inputting the sequence of the pre-training behavior data temporal span modulation propagation attenuation entropy factors into the information transfer screening module based on the gating function to obtain the sequence of the pre-training behavior data temporal span modulation propagation attenuation weights; Based on the sequence of the temporal span modulation propagation attenuation weights of the pre-training behavior data, a weighted sum of the sequence of the semantic principal component representation vectors of the historical pre-training behavior data is calculated to obtain a significant transfer aggregation representation vector of the semantic principal component of the historical pre-training behavior data; Calculate the positional sum of the semantic principal component significant transfer aggregation representation vector of the historical pre-training behavior data and the semantic principal component representation vector of the current pre-training behavior data to obtain the pre-training behavior pattern principal component temporal aggregation representation vector as the pre-training behavior pattern principal component temporal aggregation representation.

5. The personnel training data information processing method according to claim 4 is characterized in that: Calculating the absolute factors of the temporal propagation attenuation entropy of each historical pre-training behavior data semantic principal component representation vector in the sequence of the historical pre-training behavior data semantic principal component representation vector relative to the current pre-training behavior data semantic principal component representation vector to obtain a sequence of absolute factors of the temporal propagation attenuation entropy of the pre-training behavior data, including: Calculating the positional division of the historical pre-training behavior data semantic principal component representation vector and the current pre-training behavior data semantic principal component representation vector to obtain a pre-training behavior data semantic principal component attenuation vector; Calculating the logarithmic function value with base 2 of the absolute value of each eigenvalue of the semantic principal component attenuation vector of the pre-training behavior data to obtain the pre-training behavior data semantic principal component attenuation logarithmic vector; Calculate the positional point product of the semantic principal component representation vector of the historical pre-training behavior data and the attenuation logarithm vector of the semantic principal component of the pre-training behavior data, and perform positional point multiplication on the obtained point product vector to obtain the time series propagation attenuation value of the pre-training behavior data; An exponential function of the temporal propagation attenuation value of the pre-training behavior data is calculated with the natural constant e as the base to obtain the absolute factor of the temporal propagation attenuation entropy of the pre-training behavior data.

6. The personnel training data information processing method according to claim 5, characterized in that: Based on the time span between each historical pre-training behavior data semantic principal component representation vector in the sequence of historical pre-training behavior data semantic principal component representation vectors and the current pre-training behavior data semantic principal component representation vector, each pre-training behavior data temporal propagation attenuation entropy absolute factor in the sequence of pre-training behavior data temporal propagation attenuation entropy absolute factors is modulated in time dimension to obtain a sequence of pre-training behavior data temporal span modulated propagation attenuation entropy factors, including: Subtract the timestamp of the semantic principal component representation vector of the current pre-training behavior data from the timestamp of the semantic principal component representation vector of the historical pre-training behavior data and round down to obtain the time span value of the pre-training behavior data; calculate an exponential function with the natural constant e as the base and the time span value of the pre-training behavior data as the exponent to obtain the time span modulation value of the pre-training behavior data; The absolute factor of the temporal propagation attenuation entropy of the pre-training behavior data corresponding to the semantic principal component representation vector of the historical pre-training behavior data is divided by the time span modulation value of the pre-training behavior data to obtain the temporal span modulation propagation attenuation entropy factor of the pre-training behavior data.

7. The personnel training data information processing method according to claim 6 is characterized in that: Inputting the sequence of the pre-training behavior data temporal span modulation propagation attenuation entropy factors into the information transfer screening module based on the gating function to obtain the sequence of the pre-training behavior data temporal span modulation propagation attenuation weights, including: Comparing each pre-training behavior data timing span modulation propagation attenuation entropy factor in the sequence of pre-training behavior data timing span modulation propagation attenuation entropy factors with a predetermined threshold to obtain a sequence of pre-training behavior data timing span modulation propagation attenuation weights; In which, in response to the propagation attenuation entropy factor of the timing span modulation of the pre-training behavioral data being greater than the predetermined threshold, the propagation attenuation entropy factor of the timing span modulation of the pre-training behavioral data that is greater than the predetermined threshold is input into the sigmoid function; in response to the propagation attenuation entropy factor of the timing span modulation of the pre-training behavioral data being less than or equal to the predetermined threshold, the propagation attenuation entropy factor of the timing span modulation of the pre-training behavioral data that is less than or equal to the predetermined threshold is set to zero.

8. The personnel training data information processing method according to claim 7 is characterized in that: Performing behavior pattern difference semantic measurement on the principal component temporal aggregation representation of the pre-training behavior pattern and the principal component temporal aggregation representation of the post-training behavior pattern to obtain a semantic representation of the behavior pattern difference before and after training, including: inputting the principal component temporal aggregation representation vector of the pre-training behavior pattern and the principal component temporal aggregation representation vector of the post-training behavior pattern into a behavior pattern difference semantic measurement network to obtain a semantic representation vector of the behavior pattern difference before and after training as the semantic representation of the behavior pattern difference before and after training.

9. The personnel training data information processing method according to claim 8, characterized in that: Based on the semantic representation of the difference in behavior patterns before and after training, an evaluation result is obtained, including: inputting the semantic representation vector of the difference in behavior patterns before and after training into a classifier-based training effect evaluator to obtain the evaluation result, wherein the evaluation result is used to indicate whether the training effect is positive or negative.

10. A personnel training data information processing system, characterized in that: include: An employee pre-training and post-training behavior data collection module is used to obtain a time queue of pre-training behavior data and a time queue of post-training behavior data of a tracked employee object; A semantic embedding coding module for employee behavior data before and after training, used to perform semantic embedding coding on the time queue of the pre-training behavior data and the time queue of the post-training behavior data respectively to obtain a time queue of semantic embedding coding vectors of the pre-training behavior data and a time queue of semantic embedding coding vectors of the post-training behavior data; A module for temporal aggregation of behavioral data features before and after employee training, which is used to perform temporal aggregation of behavioral data features on the time queue of the semantic embedding coding vector of the pre-training behavioral data and the time queue of the semantic embedding coding vector of the post-training behavioral data to obtain a temporal aggregation representation of the principal component of the pre-training behavioral pattern and a temporal aggregation representation of the principal component of the post-training behavioral pattern, wherein the module for temporal aggregation of behavioral data features before and after employee training includes: a principal component analysis unit for behavioral data semantic embedding coding vector, which is used to extract the sequence of the current semantic embedding coding vector of the pre-training behavioral data and the sequence of the historical semantic embedding coding vector of the pre-training behavioral data and perform principal component analysis to obtain a sequence of the current semantic principal component representation of the pre-training behavioral data and the sequence of the historical semantic principal component representation of the pre-training behavioral data; a pre-training behavioral pattern principal component temporal aggregation representation generation unit, which is used to perform principal component temporal aggregation based on gradient attenuation on the sequence of the semantic principal component representation of the current semantic principal component representation of the pre-training behavioral data and the sequence of the semantic principal component representation of the historical pre-training behavioral data to obtain the principal component temporal aggregation representation of the pre-training behavioral pattern; A module for generating semantic representations of behavioral pattern differences before and after training, used for performing semantic measurement of behavioral pattern differences on the principal component temporal aggregation representation of the behavioral pattern before training and the principal component temporal aggregation representation of the behavioral pattern after training to obtain semantic representations of behavioral pattern differences before and after training; The training evaluation effect result generation module is used to obtain the evaluation result based on the semantic representation of the difference in behavior patterns before and after the training.

Images