Performance assessment method and device based on self-attention mechanism, equipment and medium

Abstract

The invention provides a performance assessment method and device based on a self-attention mechanism, equipment and a medium, and belongs to the technical field of data processing. The method comprises the following steps: in response to a performance assessment instruction, determining each piece of operation data corresponding to a to-be-assessed object in a performance assessment system; and determining first performance information of the to-be-assessed object in each job task based on each preset assessment index, each job data and a preset self-attention mechanism. According to a plurality of preset evaluation dimensions, dividing the first performance information into a plurality of evaluation data sets; determining second performance information corresponding to the to-be-assessed object based on each assessment data set and a preset self-attention mechanism; the second performance information represents a second performance contribution degree of the to-be-assessed object in the plurality of assessment dimensions. And inputting the first performance information and the second performance information into a pre-trained index adjustment prediction model to determine performance assessment performance information and an assessment index adjustment strategy corresponding to the to-be-assessed object according to a model output result.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a performance evaluation method, device, equipment and medium based on a self-attention mechanism. Background Art

[0002] Performance appraisal plays a vital role in the survival and development of an enterprise. It not only affects the daily operational efficiency of the enterprise, but also directly affects the long-term planning and competitiveness of the enterprise. Performance appraisal is not only a means of evaluating employees' past work results, but also a powerful driving force for the future development of the enterprise.

[0003] At present, the existing performance appraisal mainly measures the work performance of employees or departments through key performance indicators (KPI). KPI mainly focuses on specific and quantitative indicators and lacks humanization. In the long run, it is easy for employees or departments to sacrifice long-term interests in order to achieve indicators. KPI is also prone to human subjective judgment, which affects the fairness of the final performance appraisal, and the appraisal method is not flexible enough. Summary of the invention

[0004] The embodiments of the present application provide a performance appraisal method, apparatus, device and medium based on a self-attention mechanism, which are used to solve the technical problems that the current performance appraisal method is single and rigid, and it is difficult to provide performance appraisal results efficiently, fairly and flexibly.

[0005] On the one hand, an embodiment of the present application provides a performance evaluation method based on a self-attention mechanism, and the method is applied to a performance evaluation system; the method includes:

[0006] In response to the performance appraisal instruction, determining each operation data corresponding to the object to be appraised in the performance appraisal system;

[0007] Based on the preset assessment indicators and the operation data, and the preset self-attention mechanism, determine the first performance information of the subject to be assessed in each operation task; the first performance information represents the first performance contribution degree of the subject to be assessed under each assessment indicator;

[0008] According to a plurality of preset evaluation dimensions, each of the first performance information is divided into a plurality of evaluation data sets;

[0009] Based on each of the evaluation data sets and the preset self-attention mechanism, determining the second performance information corresponding to the object to be evaluated; the second performance information represents the second performance contribution degree of the object to be evaluated in the multiple evaluation dimensions;

[0010] The first performance information and the second performance information are input into a pre-trained indicator adjustment prediction model to determine the performance appraisal information and the appraisal indicator adjustment strategy corresponding to the object to be appraised according to the output result of the model.

[0011] In one implementation of the present application, based on the preset assessment indicators and the operation data, and the preset self-attention mechanism, the first performance information of the subject to be assessed in each operation task is determined, specifically including:

[0012] According to the preset operation task, each operation data is divided into an assessment index vector; the operation data at least includes a score of the assessment index corresponding to the object to be assessed; the same assessment index vector corresponds to the same preset operation task;

[0013] Determine a weight vector corresponding to each of the assessment indicators according to the assessment indicator vector and the preset self-attention mechanism;

[0014] The first performance information is determined according to a first dot product result of the assessment index vector and the weight vector.

[0015] In one implementation of the present application, the assessment indicators include at least: processing workload, average processing time per order, processing waiting time, order return rate, and quality inspection error rate.

[0016] In one implementation of the present application, each of the first performance information is divided into multiple evaluation data sets according to multiple preset evaluation dimensions, specifically including:

[0017] Acquire various evaluation dimensions from the user terminal; the evaluation dimensions at least include responsibility performance dimension, project contribution dimension, and teamwork dimension;

[0018] According to each of the evaluation dimensions and a pre-trained division rule model, each of the first performance information is divided into a plurality of the evaluation data sets corresponding to the number of dimensions of each of the evaluation dimensions.

[0019] In one implementation of the present application, based on each of the evaluation data sets and the preset self-attention mechanism, determining the second performance information corresponding to the object to be assessed specifically includes:

[0020] Generating evaluation vectors corresponding to each evaluation data set in chronological order;

[0021] According to the preset self-attention mechanism, calculating the similarity score corresponding to each of the evaluation vectors;

[0022] According to each of the similarity scores, calculating the attention weights corresponding to each of the evaluation dimensions;

[0023] A second dot product result is calculated between the evaluation vector of the same evaluation dimension and the corresponding attention weight, and a sum of the second dot product results is used as the second performance information.

[0024] In one implementation of the present application, the first performance information and the second performance information are input into a pre-trained indicator adjustment prediction model to determine the performance evaluation information corresponding to the object to be evaluated according to the model output result, specifically including:

[0025] The first performance information and the second performance information are processed by the indicator adjustment prediction model to determine the third performance information corresponding to the next assessment period; wherein the indicator adjustment prediction model is a pre-trained long short-term memory network LSTM model; the third performance information includes the prediction results of the first performance information and the second performance information in the next assessment period;

[0026] The first performance information, the second performance information and the third performance information are concatenated to obtain the performance evaluation performance information.

[0027] In one implementation of the present application, the performance appraisal information and appraisal indicator adjustment strategy corresponding to the object to be appraised are determined according to the model output result, specifically including:

[0028] Calculating difference information between the third performance information and the first performance information and the second performance information; the difference information is obtained based on subtraction calculation;

[0029] Matching the performance floating level corresponding to the object to be assessed according to the difference information; wherein the performance floating level is recorded in a preset performance adjustment rule list;

[0030] The assessment indicator adjustment strategy is determined according to the performance floating level; the assessment indicator adjustment strategy at least includes positive adjustment and negative adjustment.

[0031] On the other hand, the embodiment of the present application further provides a performance evaluation device based on a self-attention mechanism, the device comprising:

[0032] A first determination module is used to determine each operation data corresponding to the object to be evaluated in the performance evaluation system in response to the performance evaluation instruction;

[0033] A second determination module is used to determine the first performance information of the subject to be assessed in each task based on the preset assessment indicators and the task data and the preset self-attention mechanism; the first performance information represents the first performance contribution degree of the subject to be assessed under each assessment indicator;

[0034] A division module, used for dividing each of the first performance information into a plurality of evaluation data sets according to a plurality of preset evaluation dimensions;

[0035] A third determination module is used to determine the second performance information corresponding to the object to be assessed based on each of the evaluation data sets and the preset self-attention mechanism; the second performance information represents the second performance contribution degree of the object to be assessed in the multiple evaluation dimensions;

[0036] An input module is used to input the first performance information and the second performance information into a pre-trained indicator adjustment prediction model to determine the performance appraisal information and appraisal indicator adjustment strategy corresponding to the object to be appraised according to the model output result.

[0037] On the other hand, the embodiment of the present application further provides a performance evaluation device based on a self-attention mechanism, the device comprising:

[0038] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a performance evaluation method based on a self-attention mechanism as described above.

[0039] On the other hand, an embodiment of the present application also provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions can execute a performance evaluation method based on the self-attention mechanism as described above.

[0040] Compared with the prior art, the present invention has the following significant effects:

[0041] Through the above scheme, this application can comprehensively consider the performance contribution of the object to be evaluated under different evaluation indicators, and also consider the performance in different evaluation dimensions, and finally provide a prediction result that combines the two, and then generate performance evaluation information, and the performance evaluation method is flexible. And the evaluation indicators can be flexibly adjusted according to the object to be evaluated, so as to comprehensively evaluate the object to be evaluated and ensure the fair completion of the performance evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 A flowchart of a performance evaluation method based on a self-attention mechanism in an embodiment of the present application;

[0044] Figure 2 This is a schematic diagram of the structure of a performance evaluation device based on a self-attention mechanism in an embodiment of the present application;

[0045] Figure 3 This is a structural diagram of a performance evaluation device based on a self-attention mechanism in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0047] The embodiments of the present application provide a performance appraisal method, device, equipment and medium based on a self-attention mechanism, which are used to solve the technical problems that the current performance appraisal method is single and rigid, and it is difficult to provide performance appraisal results efficiently, fairly and flexibly.

[0048] The following describes in detail various embodiments of the present application in conjunction with the accompanying drawings.

[0049] The embodiment of the present application provides a performance evaluation method based on a self-attention mechanism, and the performance evaluation method based on a self-attention mechanism is applied to a performance evaluation system. Figure 1 As shown, the method may include steps S101-S105:

[0050] S101, the server responds to the performance appraisal instruction and determines each operation data corresponding to the object to be appraised in the performance appraisal system.

[0051] It should be noted that the server, as the executor of the performance appraisal method based on the self-attention mechanism, is only an example. The executor is not limited to the server, and this application does not make any specific limitations on this.

[0052] Performance appraisal instructions can come from user terminals, such as computers, mobile phones and other devices of appraisal department personnel, and this application does not make specific restrictions on this. The performance appraisal system can have a system platform used by roles such as appraisal subjects and appraisal department personnel. Appraisal subjects can upload work data, and appraisal department personnel can check or summarize and process data in the performance appraisal system. Among them, the object to be appraised can be a single employee or a corporate department, and this application does not make specific restrictions on this.

[0053] After receiving the performance appraisal instruction from the user terminal, the server parses the performance appraisal instruction, which carries the object to be appraised, i.e., performs performance appraisal on the object to be appraised. The server extracts the various operation data of the object to be appraised from the performance appraisal system according to the object to be appraised corresponding to the parsing result. Among them, the operation data can be understood as the completion status of the object to be appraised in the work.

[0054] For example, the workload is 20 orders per day, with a score of 80; the average processing time per order is 100 minutes, with a score of 90; the waiting time is 120 minutes, with a score of 80, etc.

[0055] S102, the server determines the first performance information of the subject to be assessed in each task based on the preset assessment indicators and operation data and the preset self-attention mechanism.

[0056] The first performance information represents the first performance contribution degree of the subject to be assessed under each assessment indicator.

[0057] In the embodiment of the present application, based on the preset assessment indicators and the operation data, and the preset self-attention mechanism, the first performance information of the subject to be assessed in each operation task is determined, specifically including:

[0058] The server divides each operation data into an assessment indicator vector according to the preset operation task. The operation data at least includes the score of the assessment indicator corresponding to the assessment object. The same assessment indicator vector corresponds to the same preset operation task. According to the assessment indicator vector and the preset self-attention mechanism, the weight vector corresponding to each assessment indicator is determined. According to the first dot product result of the assessment indicator vector and the weight vector, the first performance information is determined.

[0059] Among them, the assessment indicators include at least: processing workload, average processing time per order, processing waiting time, order return rate, and quality inspection error rate.

[0060] That is to say, the server can divide each operation data of the object to be assessed into different assessment index vectors according to the preset operation tasks. For example, each operation data of the first operation task is divided into the first assessment index vector A, and each operation data of the second operation task is divided into the second assessment index vector B. The preset operation tasks can be set by the user according to the actual usage scenario, and this application does not make specific limitations on this.

[0061] Subsequently, the server will use the preset self-attention mechanism and the assessment indicator vector to calculate the weights of different assessment indicators of the same assessment indicator vector and obtain the corresponding weight vector. The self-attention layer in the preset self-attention mechanism can use the assessment indicator vector and the pre-set weight matrix to calculate the query vector, key vector, and value vector, and then apply the normalized exponential function softmax to calculate the weight corresponding to each assessment indicator in the assessment indicator vector, and add the weights to the weight vector in sequence according to the corresponding assessment indicator arrangement order in the assessment indicator vector. Subsequently, the dot product calculation is performed based on the assessment indicator vector and the weight vector to obtain the first dot product result. The server adds the first dot product result to the first performance information.

[0062] The weight matrix is ​​set by the user according to actual use, and this application does not make any specific limitations on this.

[0063] For example, the workload is 20 orders / day, with a score of 80; the average processing time per order is 100 minutes, with a score of 90; the waiting time for processing is 120 minutes, with a score of 80; the return rate is 5%, with a score of 90; the quality inspection error rate is 2%, with a score of 95. The input assessment index vector is [80, 90, 80, 90, 95], and the weight vector calculated by the preset self-attention mechanism is [0.5, 0.2, 0.2, 0.05, 0.05]. The first dot product result is: 80*0.5+90*0.2+80*0.2+90*0.05+95*0.05=40+18+16+4.5+4.75=83.25. Here, the first performance information is the comprehensive score of an assessment index. In actual use, the calculated first dot product result can also be a vector, which is not specifically limited in this application.

[0064] S103: The server divides each first performance information into multiple evaluation data sets according to multiple preset evaluation dimensions.

[0065] In the embodiment of the present application, each first performance information is divided into multiple evaluation data sets according to multiple preset evaluation dimensions, specifically including:

[0066] The server obtains each evaluation dimension from the user terminal. The evaluation dimension includes at least a responsibility performance dimension, a project contribution dimension, and a teamwork dimension. According to each evaluation dimension and a pre-trained division rule model, each first performance information is divided into a plurality of evaluation data sets corresponding to the number of dimensions of each evaluation dimension.

[0067] In other words, the server can use a pre-trained partitioning rule model to identify the evaluation dimension corresponding to the first performance information. The partitioning rule model can be a neural network model or a machine learning model, which can distinguish the evaluation dimensions according to the source of the first performance information. In actual use, the evaluation dimensions can include responsibility performance, project contribution, and teamwork. Users can expand the evaluation dimensions based on usage. The server will divide each first performance information into multiple evaluation data sets according to the evaluation dimensions. In actual use, different job tasks may correspond to different evaluation dimensions.

[0068] Through the above division, the performance of the subject to be assessed can be determined from different dimensions, so that we are no longer bound by fixed indicators and can treat the subject to be assessed in a more humane way to complete the performance assessment.

[0069] S104: The server determines second performance information corresponding to the object to be assessed based on each evaluation data set and a preset self-attention mechanism.

[0070] The second performance information represents the second performance contribution degree of the subject to be assessed in multiple assessment dimensions.

[0071] In the embodiment of the present application, based on each evaluation data set and the preset self-attention mechanism, the second performance information corresponding to the object to be assessed is determined, specifically including:

[0072] The server generates evaluation vectors corresponding to each evaluation data set in chronological order. According to the preset self-attention mechanism, the similarity score corresponding to each evaluation vector is calculated. According to each similarity score, the attention weight corresponding to each evaluation dimension is calculated. The evaluation vector of the same evaluation dimension is used to calculate the second dot product result with the corresponding attention weight, and the sum of the second dot product results is used as the second performance information.

[0073] In other words, the server generates evaluation vectors for the first performance information in each evaluation data set in chronological order, that is, one evaluation vector corresponds to one evaluation data set, and the evaluation vector includes the first performance information in the corresponding evaluation data set arranged along the time, and one first performance information is an element in the evaluation vector. Subsequently, the server uses the above-mentioned preset self-attention mechanism to calculate the similarity score of each evaluation vector. Specifically, the evaluation vector v i represents the evaluation vector of the i-th evaluation dimension. Taking the j-th object to be evaluated as an example, through s ij =v i ·v j Calculate the similarity score, where s ij It represents the similarity score corresponding to the evaluation vector of the i-th evaluation dimension, which is the similarity score between the evaluation vector of the i-th evaluation dimension and the reference vector v j The dot product calculation result, v jIt can be understood as a reference vector, such as the weighted average of the evaluation vectors of all evaluation dimensions of the j-th object to be assessed, or a unit vector selected by the user, and this application does not make specific limitations on this.

[0074] Furthermore, the server can obtain a similarity score corresponding to the same object to be assessed in different evaluation dimensions, and the server calculates each attention weight through the following formula:

[0075]

[0076] Among them, α ij is the attention weight of the jth object to be assessed in the i-th evaluation dimension, exp represents the natural exponential function, k represents the k-th evaluation dimension, the total number of evaluation dimensions is M, the value interval of k is [1, M], and i belongs to the interval [1, M]. At this point, the server calculates an attention weight of the object to be assessed in each evaluation dimension, and then the server calculates the dot product of the attention weight and the evaluation vector of the corresponding evaluation dimension according to the corresponding relationship of the evaluation dimension to obtain the second dot product result, and then calculates the second dot product results of each evaluation dimension to obtain the second performance information, as shown in the formula h j It is the second performance information.

[0077] S105, the server inputs the first performance information and the second performance information into a pre-trained indicator adjustment prediction model to determine the performance evaluation information and the evaluation indicator adjustment strategy corresponding to the object to be evaluated according to the output result of the model.

[0078] In the embodiment of the present application, the first performance information and the second performance information are input into a pre-trained indicator adjustment prediction model to determine the performance evaluation information corresponding to the object to be evaluated according to the output result of the model, specifically including:

[0079] The first performance information and the second performance information are processed through the indicator adjustment prediction model to determine the third performance information corresponding to the next assessment cycle. Among them, the indicator adjustment prediction model is a pre-trained long short-term memory network (Long Short-Term Memory, LSTM) model. The third performance information includes the prediction results of the first performance information and the second performance information in the next assessment cycle. The first performance information, the second performance information and the third performance information are spliced ​​to obtain the performance assessment performance information.

[0080] That is to say, this application trains the LSTM model by collecting historical performance appraisal samples of several objects to be appraised, until the output accuracy of the LSTM model is greater than the accuracy threshold preset by the user, and obtains a trained LSTM model, that is, an indicator adjustment prediction model, wherein the historical performance appraisal samples include historical first performance information, historical second performance information, and historical third performance information, and the historical first performance information and historical second performance information are used as input data, and the historical third performance information is used as a training label. The server uses the indicator adjustment prediction model to make a prediction based on the first performance information and the second performance information to obtain the third performance information of the next appraisal cycle of the current appraisal cycle. The third performance information can be understood as the first performance information and the second performance information of the next appraisal cycle. The server splices the three, that is, splices the first performance information, the second performance information, and the third performance information to obtain the final performance appraisal performance information.

[0081] In actual use, the performance appraisal information can be displayed in a table. The performance appraisal information is displayed on the user terminal, and the performance appraisal information can comprehensively display multiple appraisal indicators, multiple evaluation dimensions and predicted performance of the object to be appraised, thereby comprehensively providing performance appraisal results. In addition, the present application can generate comprehensive calculation weights for the first performance information, the second performance information and the third performance information in advance, and the server can perform dot product calculations on the three with the corresponding comprehensive calculation weights to obtain a final value, which represents the performance degree of the performance appraisal of the object to be appraised.

[0082] Furthermore, the performance appraisal information and appraisal index adjustment strategy corresponding to the appraisal object are determined according to the output results of the model, including:

[0083] Calculate the difference information between the third performance information and the first performance information and the second performance information. The difference information is obtained based on subtraction. Match the performance floating level corresponding to the object to be assessed according to the difference information. The performance floating level is recorded in the preset performance adjustment rule list. Determine the assessment indicator adjustment strategy according to the performance floating level. The assessment indicator adjustment strategy includes at least positive adjustment and negative adjustment.

[0084] Specifically, calculating the difference information can be calculating the difference between the first prediction result of the first performance information and the second prediction result of the second performance information in the third performance information and the first performance information and the second performance information, respectively, to obtain a difference value or a difference vector, and to form two-dimensional difference information. The server can store a list of preset performance adjustment rules corresponding to the difference information, recording the performance floating levels corresponding to different difference information. There is a corresponding relationship between the performance floating level and the assessment index adjustment strategy. Through the above steps, the server can determine the assessment index adjustment strategy for the object to be assessed, in preparation for subsequent performance assessment. Among them, positive adjustment can be understood as increasing the index requirements, and negative adjustment can be understood as lowering the index requirements. The specific adjustment degree is set by the user according to the actual usage scenario, and this application does not make specific limitations on this.

[0085] For example, the performance floating level includes one, two, and three levels. In the difference information, the first performance information corresponds to a negative difference, and the absolute value of the difference is greater than the preset value. The second performance information corresponds to a positive difference. At this time, the performance floating level is one, and a negative adjustment is made. In the difference information, the first performance information corresponds to a negative difference, and the absolute value of the difference is not greater than the preset value. The second performance information corresponds to a positive difference. At this time, the performance floating level is two, and no assessment index adjustment is made. In the difference information, the first performance information corresponds to a positive difference, and the difference is greater than the preset value. The second performance information corresponds to a positive difference. At this time, the performance floating level is three, and a positive adjustment is made to the assessment index. The correspondence between the performance floating level and the assessment index adjustment strategy of this application can be set by the user according to the actual usage scenario, and this application does not make specific limitations on this.

[0086] In addition, the assessment indicator adjustment strategy can also include adjusting the assessment cycle and assessment method, so that performance assessment can evolve according to business development and be conducted flexibly.

[0087] Through the above scheme, this application can comprehensively consider the performance contribution of the object to be evaluated under different evaluation indicators, and also consider the performance in different evaluation dimensions, and finally provide a prediction result that combines the two, and then generate performance evaluation information, and the performance evaluation method is flexible. And the evaluation indicators can be flexibly adjusted according to the object to be evaluated, so as to comprehensively evaluate the object to be evaluated and ensure the fair completion of the performance evaluation.

[0088] Figure 2 A schematic diagram of the structure of a performance evaluation device 200 based on a self-attention mechanism provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the performance evaluation device 200 based on the self-attention mechanism includes:

[0089] The first determination module 201 is used to determine the various operation data corresponding to the object to be assessed in the performance assessment system in response to the performance assessment instruction. The second determination module 202 is used to determine the first performance information of the object to be assessed in each operation task based on the preset assessment indicators and the various operation data and the preset self-attention mechanism. The first performance information represents the first performance contribution degree of the object to be assessed under each assessment indicator. The division module 203 is used to divide each first performance information into multiple evaluation data sets according to the preset multiple evaluation dimensions. The third determination module 204 is used to determine the second performance information corresponding to the object to be assessed based on each evaluation data set and the preset self-attention mechanism. The second performance information represents the second performance contribution degree of the object to be assessed in multiple evaluation dimensions. The input module 205 is used to input the first performance information and the second performance information into a pre-trained indicator adjustment prediction model to determine the performance assessment performance information and the assessment indicator adjustment strategy corresponding to the object to be assessed according to the output result of the model.

[0090] In the embodiment of the present application, the second determining module 202 is specifically used for:

[0091] According to the preset operation tasks, each operation data is divided into an assessment indicator vector. The operation data at least includes a score of the assessment indicator corresponding to the assessment object. The same assessment indicator vector corresponds to the same preset operation task. According to the assessment indicator vector and the preset self-attention mechanism, a weight vector corresponding to each assessment indicator is determined. According to the first dot product result of the assessment indicator vector and the weight vector, the first performance information is determined.

[0092] In an embodiment of the present application, the performance evaluation device 200 based on the self-attention mechanism also includes: the evaluation indicators include at least: processing workload, average processing time per order, processing waiting time, order return rate, and quality inspection error rate.

[0093] The division module 203 is specifically used for:

[0094] Obtain each evaluation dimension from the user terminal. The evaluation dimension includes at least a responsibility performance dimension, a project contribution dimension, and a teamwork dimension. According to each evaluation dimension and a pre-trained division rule model, each first performance information is divided into a plurality of evaluation data sets corresponding to the number of dimensions of each evaluation dimension.

[0095] The third determination module 204 is specifically used for:

[0096] Generate evaluation vectors corresponding to each evaluation data set in chronological order. Calculate the similarity score corresponding to each evaluation vector according to the preset self-attention mechanism. Calculate the attention weight corresponding to each evaluation dimension according to each similarity score. Calculate the second dot product result of the evaluation vector of the same evaluation dimension and the corresponding attention weight, and use the sum of the second dot product results as the second performance information.

[0097] The input module 205 is specifically used for:

[0098] The first performance information and the second performance information are processed through the indicator adjustment prediction model to determine the third performance information corresponding to the next assessment cycle. Among them, the indicator adjustment prediction model is a pre-trained long short-term memory network LSTM model. The third performance information includes the prediction results of the first performance information and the second performance information in the next assessment cycle. The first performance information, the second performance information and the third performance information are spliced ​​to obtain the performance assessment performance information.

[0099] The input module 205 is also specifically used for:

[0100] Calculate the difference information between the third performance information and the first performance information and the second performance information. The difference information is obtained based on subtraction. Match the performance floating level corresponding to the object to be assessed according to the difference information. The performance floating level is recorded in the preset performance adjustment rule list. Determine the assessment indicator adjustment strategy according to the performance floating level. The assessment indicator adjustment strategy includes at least positive adjustment and negative adjustment.

[0101] Figure 3 A schematic diagram of the structure of a performance evaluation device based on a self-attention mechanism provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the device includes:

[0102] At least one processor; and a memory in communication with the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0103] In response to the performance appraisal instruction, determine each operation data corresponding to the object to be appraised in the performance appraisal system. Based on the preset appraisal indicators and each operation data, and the preset self-attention mechanism, determine the first performance information of the object to be appraised in each operation task. The first performance information represents the first performance contribution degree of the object to be appraised under each appraisal indicator. According to the preset multiple evaluation dimensions, divide each first performance information into multiple evaluation data sets. Based on each evaluation data set and the preset self-attention mechanism, determine the second performance information corresponding to the object to be appraised. The second performance information represents the second performance contribution degree of the object to be appraised in multiple evaluation dimensions. Input the first performance information and the second performance information into the pre-trained indicator adjustment prediction model to determine the performance appraisal performance information and the appraisal indicator adjustment strategy corresponding to the object to be appraised according to the model output result.

[0104] The embodiment of the present application further provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured as follows:

[0105] In response to the performance appraisal instruction, determine each operation data corresponding to the object to be appraised in the performance appraisal system. Based on the preset appraisal indicators and each operation data, and the preset self-attention mechanism, determine the first performance information of the object to be appraised in each operation task. The first performance information represents the first performance contribution degree of the object to be appraised under each appraisal indicator. According to the preset multiple evaluation dimensions, divide each first performance information into multiple evaluation data sets. Based on each evaluation data set and the preset self-attention mechanism, determine the second performance information corresponding to the object to be appraised. The second performance information represents the second performance contribution degree of the object to be appraised in multiple evaluation dimensions. Input the first performance information and the second performance information into the pre-trained indicator adjustment prediction model to determine the performance appraisal performance information and the appraisal indicator adjustment strategy corresponding to the object to be appraised according to the model output result.

[0106] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0107] The apparatus, equipment and medium provided in the embodiments of the present application correspond one-to-one to the method. Therefore, the apparatus, equipment and medium also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the apparatus, equipment and medium will not be repeated here.

[0108] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0109] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A performance evaluation method based on self-attention mechanism, characterized in that: The method is applied to a performance appraisal system; the method comprises: In response to the performance appraisal instruction, determining each operation data corresponding to the object to be appraised in the performance appraisal system; Based on the preset assessment indicators and the operation data, and the preset self-attention mechanism, determine the first performance information of the subject to be assessed in each operation task; the first performance information represents the first performance contribution degree of the subject to be assessed under each assessment indicator; According to a plurality of preset evaluation dimensions, each of the first performance information is divided into a plurality of evaluation data sets; Based on each of the evaluation data sets and the preset self-attention mechanism, determining the second performance information corresponding to the object to be evaluated; the second performance information represents the second performance contribution degree of the object to be evaluated in the multiple evaluation dimensions; The first performance information and the second performance information are input into a pre-trained indicator adjustment prediction model to determine the performance appraisal information and the appraisal indicator adjustment strategy corresponding to the object to be appraised according to the output result of the model.

2. A performance evaluation method based on self-attention mechanism according to claim 1, characterized in that: Based on the preset assessment indicators, the operation data, and the preset self-attention mechanism, the first performance information of the subject to be assessed in each operation task is determined, specifically including: According to the preset operation task, each operation data is divided into an assessment index vector; the operation data at least includes a score of the assessment index corresponding to the object to be assessed; the same assessment index vector corresponds to the same preset operation task; Determine a weight vector corresponding to each of the assessment indicators according to the assessment indicator vector and the preset self-attention mechanism; The first performance information is determined according to a first dot product result of the assessment index vector and the weight vector.

3. A performance evaluation method based on self-attention mechanism according to claim 2, characterized in that: The assessment indicators include at least: processing workload, average processing time per order, processing waiting time, order return rate, and quality inspection error rate.

4. The performance evaluation method based on the self-attention mechanism according to claim 1, characterized in that: According to the preset multiple evaluation dimensions, each of the first performance information is divided into multiple evaluation data sets, specifically including: Acquire various evaluation dimensions from the user terminal; the evaluation dimensions at least include responsibility performance dimension, project contribution dimension, and teamwork dimension; According to each of the evaluation dimensions and a pre-trained division rule model, each of the first performance information is divided into a plurality of the evaluation data sets corresponding to the number of dimensions of each of the evaluation dimensions.

5. A performance evaluation method based on self-attention mechanism according to claim 4, characterized in that: Based on each of the evaluation data sets and the preset self-attention mechanism, determining the second performance information corresponding to the object to be assessed specifically includes: Generating evaluation vectors corresponding to each evaluation data set in chronological order; According to the preset self-attention mechanism, calculating the similarity score corresponding to each of the evaluation vectors; According to each of the similarity scores, calculating the attention weights corresponding to each of the evaluation dimensions; A second dot product result is calculated between the evaluation vector of the same evaluation dimension and the corresponding attention weight, and a sum of the second dot product results is used as the second performance information.

6. A performance evaluation method based on self-attention mechanism according to claim 1, characterized in that: Inputting the first performance information and the second performance information into a pre-trained indicator adjustment prediction model to determine the performance evaluation information corresponding to the object to be evaluated according to the output result of the model, specifically includes: The first performance information and the second performance information are processed by the indicator adjustment prediction model to determine the third performance information corresponding to the next assessment period; wherein the indicator adjustment prediction model is a pre-trained long short-term memory network LSTM model; the third performance information includes the prediction results of the first performance information and the second performance information in the next assessment period; The first performance information, the second performance information and the third performance information are concatenated to obtain the performance evaluation performance information.

7. A performance evaluation method based on self-attention mechanism according to claim 6, characterized in that: Determine the performance appraisal information and appraisal index adjustment strategy corresponding to the object to be appraised according to the output results of the model, including: Calculating difference information between the third performance information and the first performance information and the second performance information; the difference information is obtained based on subtraction calculation; Matching the performance floating level corresponding to the object to be assessed according to the difference information; wherein the performance floating level is recorded in a preset performance adjustment rule list; The assessment indicator adjustment strategy is determined according to the performance floating level; the assessment indicator adjustment strategy at least includes positive adjustment and negative adjustment.

8. A performance evaluation device based on a self-attention mechanism, characterized in that: The device comprises: A first determination module is used to determine each operation data corresponding to the object to be evaluated in the performance evaluation system in response to the performance evaluation instruction; A second determination module is used to determine the first performance information of the subject to be assessed in each task based on the preset assessment indicators and the task data and the preset self-attention mechanism; the first performance information represents the first performance contribution degree of the subject to be assessed under each assessment indicator; A division module, used for dividing each of the first performance information into a plurality of evaluation data sets according to a plurality of preset evaluation dimensions; A third determination module is used to determine the second performance information corresponding to the object to be assessed based on each of the evaluation data sets and the preset self-attention mechanism; the second performance information represents the second performance contribution degree of the object to be assessed in the multiple evaluation dimensions; An input module is used to input the first performance information and the second performance information into a pre-trained indicator adjustment prediction model to determine the performance appraisal information and appraisal indicator adjustment strategy corresponding to the object to be appraised according to the model output result.

9. A performance evaluation device based on a self-attention mechanism, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a performance evaluation method based on a self-attention mechanism as described in any one of claims 1-7 above.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are capable of executing a performance evaluation method based on a self-attention mechanism as described in any one of claims 1 to 7 above.

Images