Talent evaluation system and evaluation method thereof

Abstract

The invention discloses a talent evaluation system, which comprises a data processing module used for extracting source data features and preprocessing the data, and outputting a data set conforming to a standard; the dynamic screening module constructs preset dimension spaces of all dimension possibilities through a rasterization permutation and combination method, and outputs a preset dimension space set; the base model module trains the models according to the training set in the current preset dimension space, and integrates the models meeting a preset threshold condition into a base model set; the combined model module firstly takes a current base model set as a first layer for training, and then selects a second layer evaluation method to generate a corresponding base model weight and a combined model group; the model evaluation module performs verification test work according to all evaluation models constructed by a preset dimension space set, and calculates evaluation indexes of the evaluation models according to the verification test work; and the multi-dimensional matching module is used for comprehensively evaluating the evaluation indexes of the evaluation models and outputting the final evaluation model and the corresponding dimensional space thereof.

Description

Technical Field

[0001] The present invention relates to the field of talent evaluation, and in particular to a talent evaluation system and an evaluation method thereof. Background Art

[0002] At present, with the rapid development of society and the growing diversified needs, various talent evaluation methods have been widely adopted and applied in different fields and scenarios, aiming to more accurately identify, evaluate and select talents. However, although these methods have achieved certain results in practice, they still inevitably have many limitations and challenges.

[0003] Specifically, some evaluation methods are easily affected by subjective factors in actual operation, such as the evaluator's personal preferences, experience, judgment, and emotional tendencies. These factors often unconsciously penetrate into the evaluation process, resulting in a highly subjective evaluation result that lacks due objectivity and fairness. This not only affects the accuracy of the evaluation, but may also lead to a series of trust issues.

[0004] In addition, it is difficult to formulate a unified and universally applicable evaluation standard when constructing some evaluation systems. Since different types of talents have significant differences in knowledge background, skill level, professional characteristics, etc., it is often necessary to redesign the entire evaluation system for different types of talents, which undoubtedly increases the complexity and cost of the evaluation work.

[0005] More importantly, some evaluation methods are technically inadequate and lack advanced data analysis techniques to effectively process and analyze massive amounts of evaluation data, making it difficult to deeply explore the value and rules hidden behind the data. This not only limits the depth and breadth of the evaluation, but also hinders the further optimization and application of the evaluation results.

[0006] Therefore, how to effectively mine the value of data, comprehensively and deeply analyze talent data, and how to screen and extract key features in multiple dimensions have become the main challenges and urgent problems faced by current research in this field. This requires not only innovation in evaluation concepts, but also continuous exploration and innovation in technical means to build a more scientific, objective and efficient talent evaluation system. Summary of the invention

[0007] The present invention is mainly to solve the above problems and provides a new talent evaluation system and evaluation method to obtain more accurate evaluation results.

[0008] The present invention adopts the following technical solutions:

[0009] A talent evaluation system comprises a data processing module, a dynamic screening module, a base model module, a combined model module, a model evaluation module and a multi-dimensional matching module; the data processing module is used for extracting source data features and preprocessing data to output a standard data set that meets the model input standard; the dynamic screening module adopts a gridding arrangement and combination method to construct a preset dimension space of all dimensional possibilities according to the selected feature dimension quantity interval range, and outputs a preset dimension space set; the base model module traverses the preset dimension space set, screens the data set according to the current preset dimension space to form a training set and a test set, performs model training based on the training set, and outputs the preset dimension space set; Models that meet the preset threshold conditions are integrated into a base model set; the combined model module first trains the current base model set as the first layer, and then selects the second-layer evaluation method to generate corresponding base model weights and combined model groups; the model evaluation module performs verification testing on all evaluation models constructed based on the preset dimensional space set, and calculates the evaluation index of each evaluation model accordingly; the multi-dimensional matching module is used to comprehensively evaluate the evaluation index of each evaluation model, and through comprehensive consideration of multiple dimensions, to determine the evaluation model with the best fit under the data set as the final evaluation model, and output this final evaluation model and its corresponding dimensional space.

[0010] Furthermore, the standard data set includes a feature dimension and a target dimension, and the target dimension includes historical evaluation scores and / or historical talent classifications.

[0011] Furthermore, the source data includes one or more talent element library sources, and the standard data set is divided into different classified data sets according to the talent element library sources, and the classified data sets become the input of the dynamic screening module.

[0012] Furthermore, the multi-dimensional matching module obtains different final evaluation models and their corresponding dimensional spaces according to the source of the talent element library.

[0013] Furthermore, when the source data includes multiple sources of talent element libraries, the talent evaluation system will train the input data and run different final evaluation models, and further output talent classification according to the highest score obtained.

[0014] Furthermore, the model training uses multiple training methods to generate multiple evaluation models.

[0015] Furthermore, the final evaluation model output value in the multi-dimensional matching module is the output value of the base model, output = w1feature1 + w2feature2 + ... + w n feature n , where w i is the weight of the feature.

[0016] Furthermore, the final evaluation model output value in the multi-dimensional matching module is the output value of the combined model, output = k1model1 + k2model2 + ... + k m model m , where k i is the weight value of the model, and k1+k2+…+k m =1.

[0017] A talent evaluation method, the method comprising the steps of:

[0018] S1, obtain source data, and continue to S2;

[0019] S2, performing data preprocessing and feature extraction on the source data, outputting a standard data set, and continuing to S3;

[0020] S3, judging whether the standard data set meets the model input rules, if not, returning to S2, if yes, continuing to S4;

[0021] S4, based on the set feature dimension quantity range, a grid arrangement and combination method is used to construct a preset dimension space of all dimensional possibilities, and a set of all preset dimension spaces is output, that is, a dimensional space data set under all possible screening conditions, and then the process continues to S5;

[0022] S5, traversing the preset dimension space set;

[0023] S6, judging whether the traversal of the preset dimension space set is completed, if the traversal is not completed, continuing to S7; if it is satisfied, outputting all evaluation models and model evaluation index values, and continuing to S10;

[0024] S7, filter out the preset dimensional space data set currently required, and use it to construct a training set and a test set, use the training set for training, so as to build a group of models, use the test data to verify each model in the model group, so as to obtain the evaluation index value of each model, and continue to S8;

[0025] S8, judging whether the evaluation index value meets the threshold condition, and counting the number of models that meet the condition; if the number of models is <= 1, returning to S5, and if the number of models is > 1, outputting the base model set, and continuing to S9;

[0026] S9, using the current base model set as the first layer for training, selecting the second layer evaluation method to generate the corresponding combination model, using the test data to verify each combination model in the combination model set to obtain the evaluation index value of each combination model, and returning to S5;

[0027] S10. Comprehensively evaluate the evaluation indicators of each evaluation model, screen out the model with the best fit through comprehensive analysis of multiple dimensions, use it as the final evaluation model, and output the final evaluation model and its corresponding dimensional space.

[0028] Furthermore, the method further comprises the steps of:

[0029] S2.5, when the source data includes multiple talent element library sources, the standard data set is divided into different classified data sets according to the talent element library sources, and each classified data set continues to S3-S10 respectively;

[0030] S11. When the source data includes multiple sources of talent element libraries, input data to train and run different final evaluation models, and further output talent classification based on the highest score obtained.

[0031] The talent evaluation system and method disclosed in the present invention are to achieve a more rapid, accurate and comprehensive evaluation of talents, so as to more effectively discover and cultivate core talents, and then promote the sustainable development and innovation process of the organization. The evaluation and analysis method uses data analysis technology to deeply explore the data information containing value, and adopts rasterization means to eliminate the influence of a large number of irrelevant features, implement multi-dimensional and dynamic feature screening, and build a unique talent evaluation model under the data set with the support of artificial intelligence technology. At the same time, the system also scientifically and rationally classifies talents into the optimal category. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The above technical content of the present invention and the following specific embodiments will be better understood when read in conjunction with the accompanying drawings. It should be noted that the accompanying drawings are only examples of the technical solutions claimed for protection. In the accompanying drawings, the same reference numerals represent the same or similar elements.

[0033] Figure 1 is a module diagram of a talent evaluation system in one embodiment of the present invention;

[0034] Figure 2 It is a flow chart of a talent evaluation method in one embodiment of the present invention. DETAILED DESCRIPTION

[0035] The detailed features and advantages of the present invention are described below in the specific implementation mode, and the content is sufficient to enable those skilled in the art to understand the technical content of the present invention and implement it accordingly. According to this specification, claims and drawings, those skilled in the art can easily understand the relevant objects and advantages of the present invention.

[0036] It should be noted that in this specification, similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0037] In the description of this embodiment, it should be noted that the orientation or position relationship indicated by the terms "upper", "lower", "inner", "bottom", etc. is based on the orientation or position relationship shown in the accompanying drawings, or is the orientation or position relationship in which the product is usually placed when used. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of this patent.

[0038] In the following, the feature dimension is the dimension composed of all the features in the element library.

[0039] The rasterization permutation and combination method refers to iterative processing for each interval value based on the specified interval value range, extracting different column data from the feature dimension, and removing duplicate items.

[0040] The preset dimensional space refers to a feature space composed of different columns of data extracted from the feature dimension.

[0041] Feature space refers to the space where feature vectors are located, where each feature corresponds to a one-dimensional coordinate axis in the feature space.

[0042] In order to make the purpose, technical solutions and advantages of this patent clearer, the implementation method of this patent will be further described in detail below in conjunction with the accompanying drawings.

[0043] The talent evaluation system disclosed in this patent includes a data processing module, a dynamic screening module, a base model module, a combination model module, a model evaluation module and a multi-dimensional matching module.

[0044] The data processing module mainly extracts features and preprocesses the collected talent data to output a data set that meets the model input standards, so as to lay the foundation for the subsequent construction of the evaluation model. The data preprocessing process, including but not limited to data cleaning, data integration, data conversion, and setting value rules for source data, aims to convert data into feature vectors suitable for artificial intelligence models. Standard data sets include feature vector space and target vector space.

[0045] Data cleaning includes processing abnormal data and initial configuration of data. Processing of abnormal data includes but is not limited to data null value processing and correction of data garbled characters. The initial setting of data may include but is not limited to: school scores are set to four levels according to school attributes, among which double first-class, 985 project, and 211 project universities are classified as level one, national key universities and full-time ordinary colleges and universities are classified as level two, private full-time ordinary colleges and universities are classified as level three, and other schools are classified as level four; professional titles are subdivided into assistant, junior, intermediate, deputy senior and senior; role scores are divided into project leaders, subject leaders, main personnel and general personnel; the source of the task talent element library is classified into national tasks, local tasks, research institute tasks and other tasks; project funding scores are divided into five intervals according to the amount: less than 100,000, less than 500,000, less than 1 million, less than 5 million, less than 10 million, less than 50 million and more than 50 million.

[0046] Data conversion includes but is not limited to data standardization and format unification, etc. For example, Excel table data is converted to CSV format.

[0047] Talent data, i.e. source data, also includes feature data, talent category data and historical evaluation scores. Feature dimensions may include but are not limited to degree category, years of work experience, party membership, graduation school score, professional title score, position, project level, project role score, achievement score, etc.

[0048] The talent element library data of source data may come from: talent element library data for undertaking major national tasks, talent element library data for basic frontier research, talent element library data for applied research and technological development, talent element library data for engineering technology, and talent element library data for scientific and technological support and guarantee.

[0049] The data processing module processes the source data and outputs the data integration result, i.e., the standard data set. The data integration result includes the feature dimension and the target dimension.

[0050] Specifically, the feature dimension is feature v =[feature1 feature2…feature d ], the target dimension includes historical talent category data and historical evaluation scores target v =[classes core];

[0051]

[0052] Among them, d is the number of features in the feature dimension; e is the number of data items in the feature dimension.

[0053]

[0054] Among them, c is the talent classification category of each data; s is the historical score of each data.

[0055] Talent categories include talents who undertake major national tasks, talents in basic frontier research, talents in applied research and technology development, talents in engineering and technology, and talents in scientific and technological support and guarantee;

[0056] Preferably, the standard data integration results are divided into different classified data sets according to different categories of talents. For example, they can be divided into data sets of talents undertaking major national tasks, data sets of talents in basic frontier research, data sets of talents in applied research and technology development, data sets of talents in engineering technology, and data sets of talents in science and technology support and guarantee.

[0057] target v .class=="NationalTaskCategory" undertakes major national tasks;

[0058] target v .class=="FrontierResearchCategory" basic frontier research category;

[0059] target v .class=="ApplicationAndTechnologyDevelopment" Application research and technology development class;

[0060] target v .class=="EngineeringTechnologyCategory"Engineering Technology Category;

[0061] target v .class=="SupportAndGuaranteeCategory" Science and technology support and guarantee category.

[0062] The dynamic screening module uses a grid-based permutation and combination method to construct a preset dimensional space of all dimensional possibilities based on the selected feature dimension quantity range, and outputs a preset dimensional space set. That is, it constructs all possible training sets and test sets, providing a solid foundation for the construction of subsequent evaluation models. In the preset dimensional space set, there are overlaps and intersections between the preset dimensional spaces. The preset dimensional space set becomes the input of the base model module. The dimension interval setting value is determined based on the user's custom input. In the talent data scenario, there are often a large number of confusing features mixed in with the effective features, affecting the judgment of truly valuable features. The dynamic screening module can minimize the impact of redundant features on the model, eliminate low-correlation or negatively correlated features, improve the model accuracy, and make the model more lightweight, and can also further reduce the requirements for data.

[0063] By using the gridding permutation and combination method, according to the set range of the number of feature dimensions, a set of all preset dimensional spaces is output, that is, a dimensional space data set under all possible screening conditions.

[0064] The feature dimension quantity interval range is [p,q], and the interval value can be p,p+1,p+2,…,q;

[0065] The set X of the preset dimensional space v =[X1X2…X g ]

[0066] Among them, g is the space number of the preset dimensional space.

[0067]

[0068] Among them, n is the feature number of the current preset dimensional space; l is the number of data items in the current preset dimensional space; j is the number of the preset dimensional space.

[0069] The base model module traverses the preset dimensional space set of the output of the dynamic screening module, and screens the data set to form a training set and a test set based on the current preset dimensional space. The base model module uses one or more training methods to perform model training based on the current preset dimensional space, as well as the training set and test set of the space, to generate one or more evaluation models. The base model module integrates the evaluation models that meet the preset threshold conditions, i.e., the base models, into a base model set. Preferably, the threshold condition can be a value or an interval. Preferably, the threshold condition is preset. Preferably, the threshold condition can be based on user-defined input. The output of the base model module is a base model set, and the base model set is also the input of the combination model module and the model evaluation module.

[0070] Training methods include but are not limited to Bayesian ridge regression algorithm, linear regression algorithm, decision tree regression algorithm, random forest regression algorithm, etc.

[0071] There is a situation where there is no basis model output, that is, in the preset dimensional space, no model that meets the preset threshold condition is found. In this case, the preset dimensional space set is traversed again and training is performed.

[0072] The data set of the preset dimensional space is randomly divided into a training set and a test set according to the ratio. The preferred ratio of the training set to the test set is 7:3.

[0073] Based on the divided test set, verify each model in the model set one by one and obtain its corresponding evaluation index value. These evaluation index values ​​include but are not limited to mean absolute error, mean square error, root mean square error and determination coefficient. Model prediction value target pre , historical evaluation score target cv .

[0074]

[0075] By comparing the evaluation index value with the threshold condition, the qualified models are screened out, and this series of qualified models are integrated into a base model set, that is, M = [Model1 Model2 … Model m ]. The base model set consists of models that meet the threshold requirements, and the base model refers to the model that meets the threshold requirements.

[0076] The combined model module first trains the current base model set as the first layer, uses the training set of the preset dimensional space to perform the first layer training and evaluates it in the process, then selects the second layer model to train the predicted value of the base model, and outputs the weights of each model in the base model set and the optimized combined model group.

[0077] The prediction score of the training results obtained by the model in the current base model set based on the training set is used as the output result of the first layer, and at the same time as the input feature space of the second layer training. In the second layer, the evaluation method is selected again to generate a combination model and its corresponding weights based on the output of the first layer, and the evaluation index value of the combination model is calculated. The combination model refers to the model that is predicted based on the established model weights after the models in the base model set are trained and their combination weights are determined.

[0078]

[0079] Among them, MP v is the input feature space for the second layer training, and K is the weight matrix.

[0080]

[0081] Among them, MA pThe predicted result is the combined model weight.

[0082] We conduct a comprehensive and detailed evaluation of the evaluation indicators of various evaluation models, and conduct in-depth analysis and comparison of the specific performance of each model in multiple dimensions. Specifically, the comparison process mainly selects the model with the best fit based on the selected indicators.

[0083] The weights of each model in the base model set and the optimized combined model become the model evaluation module.

[0084] The model evaluation module performs verification and testing on all evaluation models constructed based on the preset dimensional space set, and calculates the evaluation index of each evaluation model accordingly. Model evaluation obtains the model evaluation index value by verifying the models in the base model set and the models in the combined model group through test data. The evaluation methods used in this module include but are not limited to the holdout method, cross-validation method, and bootstrap method. Through the application of these evaluation methods, the performance of the model can be effectively judged.

[0085] Use this γ evaluation model to predict the test set to obtain the corresponding evaluation score

[0086]

[0087] According to the evaluation score, the model is evaluated using the evaluation method, where θ i is the evaluation result of each evaluation model;

[0088]

[0089] The multi-dimensional matching module is used to comprehensively evaluate the evaluation indicators of each evaluation model. By comprehensively considering multiple dimensions, it determines the evaluation model with the best fit under the data set as the final evaluation model, outputs this final evaluation model and its corresponding dimensional space, and uses this model to evaluate talents in new data.

[0090] The evaluation model with the best fit, that is, min(estimate EM ) corresponds to the evaluation model.

[0091] If the model with the best fit is the base model, the output value of the evaluation model is the output value of the base model; if the model with the best fit is the combined model, the output value of the evaluation model is the output value of the combined model.

[0092] Output values ​​of the base model:

[0093] output=w1feature1+w2feature2+…+w n feature n

[0094] Among them, w i is the weight of the feature.

[0095] The output of the combined model is:

[0096] output = k i model1+k2model2+…+k m model m

[0097] Among them, k i is the weight value of the model, and k1+k2+…+k m =1.

[0098] The above steps will be performed for the talent source data of undertaking major national tasks, talent source data of basic frontier research, talent source data of applied research and technology development, talent source data of engineering and technology, and talent source data of scientific and technological support and guarantee to obtain the output score value of the corresponding optimal model. Finally, according to the talent category corresponding to the highest score value obtained, talents can be classified into talents undertaking national tasks, talents in basic frontier research, talents in applied research and technology development, talents in engineering and technology, and talents in scientific and technological support and guarantee.

[0099] The talent evaluation method disclosed in the present invention comprises the following steps:

[0100] S1, obtain source data, and continue to S2;

[0101] S2, performing data preprocessing and feature extraction on the source data, outputting a standard data set, and continuing to S3;

[0102] S2.5, when the source data includes multiple talent element library sources, the standard data set is divided into different classified data sets according to the talent element library sources, and each classified data set continues to S3-S10 respectively;

[0103] S3, judging whether the standard data set meets the model input rules, if not, returning to S2, if yes, continuing to S4;

[0104] S4, based on the set feature dimension quantity range, a grid arrangement and combination method is used to construct a preset dimension space of all dimensional possibilities, and a set of all preset dimension spaces is output, that is, a dimensional space data set under all possible screening conditions, and then the process continues to S5;

[0105] S5, traversing the preset dimension space set;

[0106] S6, judging whether the traversal of the preset dimension space set is completed, if the traversal is not completed, continuing to S7; if it is satisfied, outputting all evaluation models and model evaluation index values, and continuing to S10;

[0107] S7, filter out the preset dimensional space data set currently required, and use it to construct a training set and a test set, use the training set for training, so as to build a group of models, use the test data to verify each model in the model group, so as to obtain the evaluation index value of each model, and continue to S8;

[0108] S8, judging whether the evaluation index value meets the threshold condition, and counting the number of models that meet the condition; if the number of models is <= 1, returning to S5, and if the number of models is > 1, outputting the base model set, and continuing to S9;

[0109] S9, using the current base model set as the first layer for training, selecting the second layer evaluation method to generate the corresponding combination model, using the test data to verify each combination model in the combination model set to obtain the evaluation index value of each combination model, and returning to S5;

[0110] S10. Comprehensively evaluate the evaluation indicators of each evaluation model, screen out the model with the best fit through comprehensive analysis of multiple dimensions, use it as the final evaluation model, and output the final evaluation model and its corresponding dimensional space.

[0111] S11. When the source data includes multiple sources of talent element libraries, the input data runs different final evaluation models, and talent classification is further output based on the highest score obtained.

[0112] This talent evaluation and analysis method is designed to achieve a faster, more accurate and comprehensive evaluation of talents, so as to more effectively discover and cultivate core talents, and thus promote the sustainable development and innovation process of the organization. This evaluation and analysis method uses data analysis technology to deeply explore valuable data information and data relationships, adopts rasterization methods, implements multi-dimensional and dynamic feature screening, and builds a unique talent evaluation model under this data set with the support of artificial intelligence technology. At the same time, the system also scientifically and rationally classifies talents into the optimal category.

[0113] Compared with data analysis systems that have not been model-compared or rasterized, this system can more deeply explore the spatial dependencies and intrinsic structural characteristics between data by integrating the input data after rasterization, which is crucial to improving the accuracy of analysis. In addition, by using multiple models in parallel for prediction or classification, and combining the output of each model with a weighted strategy, this system can make full use of the advantages of different models in specific fields and avoid possible limitations, thereby achieving significant improvements in prediction accuracy, robustness, and the ability to handle complex problems compared to the comparison system. This improvement is reflected in the quantitative evaluation as a better fit between the prediction and historical scores, and the variance between the prediction and historical scores is reduced by 15%. And in the face of changes in internal structure or external environment, it can still maintain its ability to operate stably.

[0114] The terms and expressions used herein are for descriptive purposes only, and the present invention should not be limited to these terms and expressions. The use of these terms and expressions does not mean to exclude any equivalent features of the illustrations and descriptions (or parts thereof), and it should be recognized that various modifications that may exist should also be included in the scope of the claims. Other modifications, variations and substitutions may also exist. Accordingly, the claims should be deemed to cover all such equivalents.

[0115] Similarly, it should be pointed out that although the present invention has been described with reference to the current specific embodiments, ordinary technicians in this technical field should realize that the above embodiments are only used to illustrate the present invention, and various equivalent changes or substitutions may be made without departing from the spirit of the present invention. Therefore, as long as the changes and modifications to the above embodiments are within the scope of the essential spirit of the present invention, they will fall within the scope of the claims of the present invention.

Claims

1. A talent evaluation system, characterized in that: It includes data processing module, dynamic screening module, base model module, combination model module, model evaluation module and multi-dimensional matching module; The data processing module is used to extract source data features and preprocess the data to output a standard data set that meets the model input standard; The dynamic screening module constructs a preset dimensional space of all dimensional possibilities based on the selected feature dimension quantity interval range, and outputs a preset dimensional space set by using a grid arrangement and combination method; The base model module traverses the preset dimensional space set, screens the data set to form a training set and a test set according to the current preset dimensional space, performs model training based on the training set, and integrates the models that meet the preset threshold conditions into the base model set; The combined model module first trains the current base model set as the first layer, and then selects the second layer evaluation method to generate the corresponding base model weights and combined model group; The model evaluation module performs verification testing on all evaluation models constructed based on the preset dimensional space set, and calculates the evaluation index of each evaluation model accordingly; The multi-dimensional matching module is used to comprehensively evaluate the evaluation indicators of each evaluation model, and to determine the evaluation model with the best fit under the data set as the final evaluation model by comprehensively considering multiple dimensions, and output the final evaluation model and its corresponding dimensional space.

2. The talent evaluation system according to claim 1, characterized in that: The standard data set includes a feature dimension and a target dimension, and the target dimension includes historical evaluation scores and / or historical talent classifications.

3. The talent evaluation system according to claim 2, characterized in that: The source data includes one or more talent element library sources, and the standard data set is divided into different classified data sets according to the talent element library sources, and the classified data sets become the input of the dynamic screening module.

4. The talent evaluation system according to claim 3, characterized in that: The multi-dimensional matching module obtains different final evaluation models and their corresponding dimensional spaces according to the source of the talent element library.

5. The talent evaluation system according to claim 3, characterized in that: When the source data includes multiple sources of talent element libraries, the talent evaluation system will train the input data and run different final evaluation models, and further output talent classification according to the highest score obtained.

6. The talent evaluation system according to claim 1, characterized in that: The model training uses a variety of training methods to generate a variety of evaluation models.

7. The talent evaluation system according to claim 1, characterized in that: The final evaluation model output value in the multi-dimensional matching module is the output value of the base model, output=model=w1feature1+w2feature2+…+w n feature n where w i is the weight of the feature.

8. The talent evaluation system according to claim 1, characterized in that: The final evaluation model output value in the multi-dimensional matching module is the output value of the combined model. output=k1model1+k2model2+…+k m model m where k i is the weight value of each model in the combined model, and k1+k2+…+k m =1.

9. A talent evaluation method, characterized in that: The method comprises the steps of: S1, obtain source data, and continue to S2; S2, performing data preprocessing and feature extraction on the source data, outputting a standard data set, and continuing to S3; S3, judging whether the standard data set meets the model input rules, if not, returning to S2, if yes, continuing to S4; S4, based on the set feature dimension quantity interval range, a grid arrangement and combination method is used to construct a preset dimensional space of all dimensional possibilities, and a set of all preset dimensional spaces is output, that is, a dimensional space data set under all possible screening conditions, and then continue to S5; S5, traversing the preset dimension space set; S6, judging whether the traversal of the preset dimension space set is completed, if the traversal is not completed, continuing to S7; if it is satisfied, outputting all evaluation models and model evaluation index values, and continuing to S10; S7, filter out the preset dimensional space data set currently required, and use it to construct a training set and a test set, use the training set for training, so as to build a group of models, use the test data to verify each model in the model group, so as to obtain the evaluation index value of each model, and continue to S8; S8, judging whether the evaluation index value meets the threshold condition, and counting the number of models that meet the condition; if the number of models is <= 1, returning to S5, and if the number of models is > 1, outputting the base model set, and continuing to S9; S9, using the current base model set as the first layer for training, selecting the second layer evaluation method to generate the corresponding combination model, using the test data to verify each combination model in the combination model set to obtain the evaluation index value of each combination model, and returning to S5; S10. Comprehensively evaluate the evaluation indicators of each evaluation model, screen out the model with the best fit through comprehensive analysis of multiple dimensions, use it as the final evaluation model, and output the final evaluation model and its corresponding dimensional space.

10. The talent evaluation method according to claim 9, characterized in that: Also includes the steps: S2.

5. When the source data includes multiple talent element library sources, the standard data set is divided into different classified data sets according to the talent element library sources, and each classified data set continues to S2-S10 respectively; S11. When the source data includes multiple sources of talent element libraries, input data to train and run different final evaluation models, and further output talent classification based on the highest score obtained.

Images