CN-122022562-A - Five-dimensional dynamic talent portrait intelligent decision management system integrating large models

CN122022562ACN 122022562 ACN122022562 ACN 122022562ACN-122022562-A

Abstract

The invention relates to the technical field of intelligent management of human resources, in particular to a five-dimensional dynamic talent portrait intelligent decision-making management system integrating a large model. The system comprises a five-dimensional data acquisition module, an image construction module, an intelligent decision module and a management optimization module, wherein the five-dimensional data acquisition module is used for acquiring a multi-source data set, the multi-source data set comprises manpower feature data, scientific research innovation feature data, ecological cooperation feature data, organization operation feature data and system execution feature data, and the image construction module is used for carrying out normalization processing and weight distribution on the multi-source data set to generate a fusion evaluation result. The invention realizes accurate construction of talent portrait based on human resource intelligent management technology, full-flow collaborative optimization for breeding and dynamic short-board identification, and greatly improves the accuracy, efficiency and suitability of talent management.

Inventors

JIANG CHUN
WEI CHENXIN
JIANG ZHENGQING
Gao Luxiong
HU GUOSHAN
JIA ZHIWEI
FANG HAOTIAN
CHEN JING
DU SIYUAN
XIE CHUANQI
HOU CHUN
FENG CHUANYONG
DENG SHAN
LUO XING
WANG JING
LIANG QIYUN
LI JIE
Yuan Xiongyan

Assignees

长江水利委员会水文局

Dates

Publication Date: 20260512
Application Date: 20260113

Claims (10)

1. A five-dimensional dynamic talent portrait intelligent decision management system integrating a large model is characterized by comprising the following modules: the five-dimensional data acquisition module is used for acquiring a multi-source data set, wherein the multi-source data set comprises manpower characteristic data, scientific research innovation characteristic data, ecological cooperation characteristic data, organization operation characteristic data and system execution characteristic data; the image construction module is used for carrying out normalization processing and weight distribution on the multi-source data set to generate a fusion evaluation result; The intelligent decision module is used for acquiring post demand information, inputting the post demand information and the five-dimensional talent portrait into a pre-trained fusion large model, calculating the fitness, and outputting a selected fitness score; and the management optimization module is used for carrying out quantitative analysis and associated calculation according to the multi-source data set to generate collaborative short-plate data, and generating a talent cultivation path report according to the collaborative short-plate data and the selected fit score.
2. The large-model-fused five-dimensional dynamic talent portrait intelligent decision management system according to claim 1, further comprising, after generating a talent cultivation path report according to the collaborative short-panel data and the election fitness score: the talent distribution data is mapped to a preset building information model in a correlated way so as to output a five-dimensional image intelligent visual report; setting four-level authority of the system according to the five-dimensional image intelligent visual report, wherein the four-level authority comprises manager authority, HR special personnel authority, department responsible person authority and employee authority; And encrypting and transmitting the four-level authority and the five-dimensional portrait intelligent visual report to a storage end, executing intelligent desensitization processing on the five-dimensional portrait intelligent visual report by the storage end, and automatically recording an operation log of the intelligent desensitization processing.
3. The large-model-fused five-dimensional dynamic talent representation intelligent decision management system according to claim 1, wherein the five-dimensional data acquisition module comprises the following functions: Performing image-text analysis on a preset human resource standard data set by using an ORC (object oriented computer) identification technology to extract human characteristic data, accessing an external scientific research management system to extract scientific research innovation characteristic data, collecting form original data and constructing message collaborative interaction characteristics; collecting operation data of a qualification structure, carrying out time serialization processing to obtain organization operation characteristic data, importing a full-chain system file, extracting system clause elements through semantic analysis, and associating corresponding execution records to obtain system execution characteristic data, wherein a multi-source data set comprises manpower characteristic data, scientific research innovation characteristic data, ecological cooperation characteristic data, organization operation characteristic data and system execution characteristic data as a multi-source data set.
4. The five-dimensional dynamic talent portrayal intelligent decision management system fusing a large model of claim 1, wherein the portrayal construction module comprises the following functions: The method comprises the steps of carrying out normalization processing on a multi-source data set to obtain a multi-source data set to be processed, setting five-dimensional weight based on the multi-source data set to be processed, and calculating a five-dimensional comprehensive evaluation value according to the five-dimensional weight; And calculating a fusion evaluation result by using the five-dimensional comprehensive evaluation value and a preset evaluation threshold, wherein the error of the fusion evaluation result is controlled to be +/-2%, and generating a five-dimensional talent image according to the fusion evaluation result.
5. The large-model-fused five-dimensional dynamic talent image intelligent decision management system of claim 4, wherein generating a five-dimensional talent image according to the fusion evaluation result comprises: Constructing a five-dimensional talent image comprising a personal capability base, a scientific potential, an ecological adaptation degree, an organization energized matching degree and a system adaptation degree according to the fusion evaluation result; And if the fusion evaluation result is updated, automatically writing the portrait label into the five-dimensional talent portrait.
6. The five-dimensional dynamic talent portrait intelligent decision management system fusing a large model according to claim 1, wherein the intelligent decision module comprises the following functions: The post demand information and the five-dimensional talent portrait are input into a pre-trained fusion large model, the post demand information is pre-trained, fine adjustment training is carried out according to preset organization history matching data, so as to calculate the adaptation degree, and a selective adaptation degree score is output; The fusion large model is based on a preset open source large model, and an attention mechanism module is introduced into a neural network hidden layer so as to automatically focus five-dimensional features in a multi-source data set, and a traditional Sigmoid function in the open source large model is replaced by an ELU function.
7. The large model fused five-dimensional dynamic talent representation intelligent decision management system of claim 6, wherein pre-training the post demand information and performing small sample fine-tuning training according to preset organization history matching data to perform fitness calculation, and outputting a pull-out fitness score comprises: Performing unsupervised pre-training on the word segmentation information by adopting a fusion large model, and extracting core requirement information; setting the fine-tuning learning rate of the fusion large model to be 1/10 of the learning rate in the unsupervised pre-training according to the matching label; The method comprises the steps of inputting core demand information into a fusion large model, outputting post side vectors, calculating candidate side vectors according to the post demand information, calculating cosine similarity according to the post side vectors and the candidate side vectors, and grading the cosine similarity as a selection adaptation degree.
8. The large-model-fused five-dimensional dynamic talent portrait intelligent decision management system according to claim 6, wherein the large-model-fused is based on a preset open-source large model and introduces an attention mechanism module in a neural network hidden layer, and the system comprises: the fusion large model is based on a preset open source large model and introduces an attention mechanism module in a neural network hidden layer, wherein the attention mechanism module comprises a self-attention layer and a multi-head attention layer; the self-attention layer is used for calculating the interdependence relation of five-dimensional talent portraits, the multi-head attention layer is used for executing a plurality of self-attention calculations, splicing and linearly transforming the interdependence relation, and the head number of the attention mechanism module is set to be 8 or 16.
9. The large-model-fused five-dimensional dynamic talent portrait intelligent decision management system according to claim 1, wherein the management optimization module includes the following functions: Calculating a correlation coefficient according to the multi-source data set, and if the correlation coefficient is more than or equal to 0.6, judging that the correlation coefficient is less than or equal to 0.3, judging that the correlation coefficient is less than or equal to 0.6, judging that the correlation coefficient is medium, and judging that the correlation coefficient is less than or equal to 0.3; The method comprises the steps of selecting a multi-source data set, wherein strong association, medium association and weak association are coupling relations of any two dimension data sets in the multi-source data set, identifying cooperative short plates according to the coupling relations to generate cooperative short plate data, and generating talent cultivation path reports according to the cooperative short plate data and the selected fit scores.
10. The large model fused five-dimensional dynamic talent portrait intelligent decision management system of claim 9, wherein identifying collaborative short boards according to coupling relationships, generating collaborative short board data comprises: The method comprises the steps of extracting each dimension score in a multi-source data set, setting each dimension quantization threshold value, comparing each dimension score with each dimension quantization threshold value, identifying single-dimension substandard features, and judging a Shan Weiduan plate when the single-dimension substandard features are identified and the coupling relationship is weak; Comparing the dimension scores with the dimension quantization thresholds, and judging to be a linkage weak short plate if any two dimension scores reach standards and the coupling relation is weak association; shan Weiduan plates, ganged weak shortplates, and key conductive shortplates were integrated into collaborative shortplate data.

Description

Five-dimensional dynamic talent portrait intelligent decision management system integrating large models Technical Field The invention relates to the technical field of intelligent management of human resources, in particular to a five-dimensional dynamic talent portrait intelligent decision-making management system integrating a large model. Background In the existing talent management system, the talent state is generally recorded and analyzed in a mode based on static data acquisition or preset rules, but in an application environment with complex organization structure, multiple innovation activities or multidimensional talent development paths, the traditional talent management system is difficult to comprehensively reflect talent growth and organization operation states. The existing system takes personal basic information, post histories or performance data as a core, lacks unified acquisition and association modeling of multidimensional factors such as personal capability evolution, scientific research innovation participation, ecological cooperation, organization matching, system environment and the like, is difficult to support multi-factor collaborative analysis, simultaneously, multiple independent operation of modules such as selection, cultivation, management and use and the like, limited data interaction among systems, lacks a linkage mechanism based on a unified data model, mainly comprises staged data records, lacks continuous monitoring and dynamic parameter updating capability of talent state change processes, is difficult to adapt to dynamic adjustment of organization development rhythm and talent growth period, and is insufficient in association analysis capability of system execution data and talent state data, and closed-loop operation of full period management cannot be realized. The technical limitations cause the existing talent management system to have defects in the aspects of data coverage breadth, module coordination degree and dynamic adaptability, and the requirements of the talent management system in a complex organization environment are difficult to meet. Disclosure of Invention Based on the above, the present invention is necessary to provide a five-dimensional dynamic talent portrait intelligent decision management system integrating large models, so as to solve at least one of the above technical problems. In order to achieve the purpose, the five-dimensional dynamic talent portrait intelligent decision-making management system integrating the large model comprises the following modules: the five-dimensional data acquisition module is used for acquiring a multi-source data set, wherein the multi-source data set comprises manpower characteristic data, scientific research innovation characteristic data, ecological cooperation characteristic data, organization operation characteristic data and system execution characteristic data; the image construction module is used for carrying out normalization processing and weight distribution on the multi-source data set to generate a fusion evaluation result; The intelligent decision module is used for acquiring post demand information, inputting the post demand information and the five-dimensional talent portrait into a pre-trained fusion large model, calculating the fitness, and outputting a selected fitness score; and the management optimization module is used for carrying out quantitative analysis and associated calculation according to the multi-source data set to generate collaborative short-plate data, and generating a talent cultivation path report according to the collaborative short-plate data and the selected fit score. The beneficial effects of the invention are as follows: (1) The dimension full coverage and intelligent analysis is that the system combines a large model with five-factor logic depth, and realizes comprehensive quantitative characterization of talent core elements by carrying out automatic acquisition, normalization processing and weighted fusion on five-dimension data of individuals, traumatology, ecology, organization and systems, thereby solving the problems of single dimension and data splitting of the traditional system. (2) And the decision is accurate, a large model drives a multi-source data fusion and calculation algorithm, so that the personnel selection adaptation degree scoring output is realized, the selection accuracy is more than or equal to 95%, the personnel post adaptation degree is improved by about 40% compared with the traditional method, and the manual subjective intervention is obviously reduced. (3) And the process collaboration is realized by the system, through the large model, the data links of all links are selected, cultivated, used and reserved, so that the whole process calculation, the association analysis and the result generation are realized, and the management efficiency is improved by about 50%. (4) The large model supports real-time acquisition and calculation upda