Search

CN-121981870-A - Public policy social influence assessment and dynamic adaptation method and system thereof

CN121981870ACN 121981870 ACN121981870 ACN 121981870ACN-121981870-A

Abstract

The invention provides a public policy social influence assessment and dynamic adaptation method and a system thereof, belonging to the technical field of public policy analysis and intelligent decision making, wherein the method comprises a multi-source data acquisition step for acquiring multi-source heterogeneous policy data such as government service, statistical monitoring, public opinion monitoring, citizen hot line complaint, third party investigation and the like; the method comprises the steps of data fusion preprocessing, entity alignment and timestamp normalization, influence index system construction, construction of an economic benefit, social fairness, public satisfaction and execution efficiency four-dimensional index system, emotion analysis, fine granularity emotion recognition and theme clustering by adopting a pre-training language model, comprehensive evaluation, comprehensive grading by adopting a heterogeneous graph neural network to fuse quantitative indexes and qualitative feedback, and dynamic adaptation, wherein the identification of execution deviation and negative effects generates adaptation suggestions and forms closed-loop feedback.

Inventors

  • Bu Naiqing

Assignees

  • 三亚学院

Dates

Publication Date
20260505
Application Date
20260212

Claims (10)

  1. 1. The public policy social influence assessment and dynamic adaptation method is characterized by comprising the following steps: a multi-source data acquisition step of acquiring a multi-source heterogeneous policy data set comprising government service data, statistical monitoring data, public opinion monitoring data, citizen hot line complaint data and third party investigation data; a data fusion preprocessing step, namely performing entity alignment and timestamp normalization processing on the multi-source heterogeneous policy data set to generate a fused policy data set; An influence index system construction step of constructing a policy influence assessment index system covering an economic benefit index value, a social fairness index value, a public satisfaction index value and an execution efficiency index value based on the fused policy data set; Carrying out fine granularity emotion recognition on the public opinion text and complaint content in the fused policy data set by adopting a pre-training language model, and executing topic clustering processing to generate a fine granularity emotion recognition result and a topic clustering result; A comprehensive evaluation step, in which a heterogeneous graph neural network is adopted to fuse the policy influence evaluation index system with the fine granularity emotion recognition result, and a policy comprehensive influence score and a itemized diagnosis result are output; And a dynamic adaptation step of identifying policy execution deviation information and negative effect information according to the policy comprehensive influence score and the item diagnosis result to generate a policy adaptation suggestion, wherein the policy adaptation suggestion is fed back to the multi-source data acquisition step for updating data monitoring key points to form a closed loop for policy evaluation and adaptation.
  2. 2. The method of claim 1, wherein in the multi-source data collection step, the government service data comprises administrative approval transaction amount data, public service coverage rate data and policy execution progress data, the statistical monitoring data comprises economic development index data, folk life guarantee index data and environment quality index data, the public opinion monitoring data comprises social media discussion data, news media report data and internet forum comment data, and the citizen hot line complaint data comprises complaint classification data, processing timeliness data and satisfaction evaluation data.
  3. 3. The method of claim 1, wherein the entity alignment process comprises building a policy entity knowledge graph, calculating cross-source entity similarity using an embedded entity alignment algorithm, and performing entity merging when the cross-source entity similarity exceeds a preset entity alignment threshold, wherein the timestamp normalization process comprises unifying time granularity of different data sources to a preset time unit, and assigning a time sequence weight to historical data using a time decay function.
  4. 4. The method of claim 1, wherein the calculating of the economic benefit index value comprises collecting economic output data before and after policy implementation and calculating a policy marginal economic contribution rate, the calculating of the social fairness index value comprises collecting benefit group distribution data and calculating a policy benefit face balance index, the calculating of the public satisfaction index value comprises collecting poll data and public opinion emotion data and calculating a weighted satisfaction score, and the calculating of the execution efficiency index value comprises collecting policy execution cycle data and resource input data and calculating a unit input-output efficiency value.
  5. 5. The method of claim 1, wherein the pre-training language model employs a domain-adaptive tuning strategy comprising continuously pre-training the pre-training language model using a public policy domain corpus and performing supervised tuning using a policy emotion markup dataset, and wherein the fine-grained emotion recognition comprises recognizing emotion polarity categories, emotion intensity levels, and emotion trigger objects.
  6. 6. The method of claim 1, wherein the topic clustering process employs a mixed clustering strategy combining hierarchical clustering and density clustering, and the mixed clustering strategy comprises identifying core topic clusters by adopting density clustering, and then subdividing and classifying boundary topics by adopting hierarchical clustering to generate topic clustering results with a hierarchical structure.
  7. 7. The method of claim 1, wherein the constructing of the heterogeneous graph neural network comprises constructing a policy influence evaluation heterogram by taking a policy entity, a data source, an evaluation index and a public opinion topic as heterogeneous nodes and taking an entity association relationship, a data attribution relationship and a topic correlation relationship as heterogeneous edges, and the fusing of the heterogeneous graph neural network comprises aggregating heterogeneous neighbor information by adopting a graph meaning force mechanism based on a meta path to generate a policy influence comprehensive representation vector.
  8. 8. The method of claim 1, wherein the identifying of the policy-implemented deviation information includes comparing the sub-term diagnosis result with a preset policy target value, calculating a degree of deviation of each dimension, and generating deviation pre-warning information when the degree of deviation exceeds a preset deviation threshold, and wherein the identifying of the negative effect information includes detecting a negative emotion aggregate region based on the fine-grained emotion recognition result, and locating a negative effect specific type in combination with the topic clustering result.
  9. 9. The method of claim 1, wherein the generating of the policy adaptation advice includes constructing a bias-adaptation knowledge base based on policy execution bias information, retrieving historical adaptation schemes in similar bias scenarios using a case-based reasoning approach, and generating parameter adjustment advice or corollary measure optimization advice in combination with current policy context, wherein the feeding back of the policy adaptation advice to the multi-source data collection step includes dynamically adjusting data collection frequency and monitoring index weight according to bias field identification information in the policy adaptation advice, wherein the bias field identification information includes an evaluation dimension type identification that the bias degree exceeds a preset bias threshold, and wherein the adjusting of the data collection frequency includes shortening a collection period of bias field corresponding data sources to one half of an original collection period.
  10. 10. A public policy social impact assessment and dynamic adaptation system for implementing the method of any of claims 1-9, comprising: the multi-source data acquisition module is used for acquiring a multi-source heterogeneous policy data set comprising government service data, statistical monitoring data, public opinion monitoring data, citizen hot line complaint data and third party investigation data; the data fusion preprocessing module is used for carrying out entity alignment and time stamp normalization processing on the multi-source heterogeneous policy data set to generate a fused policy data set; The influence index system construction module is used for constructing a policy influence assessment index system covering an economic benefit index value, a social fairness index value, a public satisfaction index value and an execution efficiency index value based on the fused policy data set; The emotion analysis module is used for carrying out fine granularity emotion recognition on the public opinion text and the complaint content in the fused policy data set by adopting a pre-training language model, and carrying out topic clustering processing to generate a fine granularity emotion recognition result and a topic clustering result; The comprehensive evaluation module is used for fusing the policy influence evaluation index system and the fine granularity emotion recognition result by adopting a heterogeneous graph neural network and outputting a policy comprehensive influence score and a itemized diagnosis result; The dynamic adaptation module is used for identifying policy execution deviation information and negative effect information according to the policy comprehensive influence score and the item diagnosis result to generate policy adaptation suggestions, wherein the dynamic adaptation module is in communication connection with the multi-source data acquisition module, and feeds the policy adaptation suggestions back to the multi-source data acquisition module for updating data monitoring key points.

Description

Public policy social influence assessment and dynamic adaptation method and system thereof Technical Field The invention belongs to the technical field of public policy analysis and intelligent decision making, and particularly relates to a public policy social influence assessment and dynamic adaptation method and a system thereof. Background Public policies act as a central tool for governmental governments that make and enforce levels of public welfare and efficiency that directly affect social resource allocation. With the deep advancement of modern processes of national governance systems and governance capabilities, the establishment of scientific, systematic and dynamic public policy assessment mechanisms has become a key element for improving government decision quality. However, the traditional policy evaluation method mainly relies on manual investigation and expert evaluation, has the inherent defects of long evaluation period, limited coverage, strong subjectivity and the like, and is difficult to meet the urgent requirement of modern public treatment on real-time feedback of policy effects. The invention CN109766416A discloses a new energy policy information extraction method and a system, the method processes the new energy policy into a new energy policy text with a set format, based on a pre-established new energy policy feature dictionary library, adopts a method of automatically identifying clauses by punctuation marks to carry out sentence hierarchy division and marking on the new energy policy text, outputs division and marking results to generate a text sentence library, and finally generates key information of the new energy policy feature dictionary library and the text sentence library according to the text sentences based on a unified service interface and outputs the key information. According to the method, the weight of the feature words is calculated through a TF-IDF algorithm, and the LDA topic model is utilized to conduct topic analysis, so that automatic analysis of the new energy policy text and key index extraction are achieved. However, the method has the technical limitations that firstly, the data source is single, analysis is only carried out on the policy text, multi-source heterogeneous information such as government service data, public opinion monitoring data, citizen complaint data and the like cannot be integrated, so that the view angle of the policy effect evaluation is fragmented, secondly, the analysis method is limited to word frequency statistics and topic clustering, fine granularity recognition capability on public emotion tendencies is lacking, a folk feedback signal in the policy implementation process cannot be accurately captured, thirdly, the evaluation result only outputs key information and classification labels, a systematic policy influence index system cannot be established, the comprehensive performance of the policy in multiple dimensions such as economic benefit, social fairness, public satisfaction and the like is difficult to quantify, fourthly, a closed loop feedback mechanism for policy adaptation is lacking, the evaluation result cannot be directly converted into policy optimization suggestions, and the recognition and correction of policy execution deviation still needs to rely on manual judgment. With the rapid development of big data technology and artificial intelligence algorithms, data driven policy evaluation paradigms are emerging. Extensive policy research efforts based on text mining, emotion analysis and machine learning have been conducted in academia and industry. However, existing research is mostly focused on a single policy domain or specific analysis task, lacking a systematic solution for public policy full cycle effect monitoring. Particularly, in key technical links such as multi-source heterogeneous data fusion, quantitative index and qualitative feedback collaborative analysis, policy execution deviation automatic identification and adaptive suggestion generation, significant technical blank still exists. Therefore, development of a public policy social impact assessment method and system capable of fusing multiple data, constructing comprehensive assessment indexes and realizing intelligent diagnosis and dynamic adaptation is needed, and technical support is provided for evidence-based decision and policy optimization. Disclosure of Invention Aiming at the technical problems that public policy evaluation data is single in source, evaluation indexes are incomplete, a dynamic adaptation mechanism is lacked and the like in the prior art, the invention provides a public policy social influence evaluation and dynamic adaptation method and a system thereof. In order to achieve the above purpose, the invention adopts the following technical scheme: A public policy social influence assessment and dynamic adaptation method comprises the steps of collecting multisource data, collecting multisource heterogeneous policy data se