CN-120781938-B - Digital treatment method and system for basic society

CN120781938BCN 120781938 BCN120781938 BCN 120781938BCN-120781938-B

Abstract

The invention discloses a digital treatment method and system for a basic society, wherein the method comprises the steps of collecting multi-source heterogeneous data in real time through an Internet of things sensing network deployed in a basic community to generate a standardized multi-mode sensing data stream, outputting a time-space correlated cleaned data set based on the multi-mode sensing data stream, inputting the cleaned data set into a multi-scale space-time encoder to generate a characteristic tensor comprising regional hot spot distribution and a risk propagation path, constructing a dynamic risk knowledge graph based on the characteristic tensor, outputting a decision matrix comprising a risk level and an optimal intervention path, inputting the decision matrix into a strategy optimization engine to obtain a layered treatment instruction set, and automatically correcting conflict instructions in a strategy chain through a fuzzy reinforcement learning algorithm based on real-time treatment feedback data after the layered treatment instruction set is executed. By utilizing the embodiment of the invention, a digital treatment scheme which is efficient, intelligent and has traceability can be realized, so that the social treatment efficiency and the service quality are improved.

Inventors

WANG HUI
WU ZHICHENG
ZHOU YEFEI

Assignees

浙江省邮电工程建设有限公司

Dates

Publication Date: 20260508
Application Date: 20250619

Claims (9)

1. A method for digitally managing a base society, the method comprising: Acquiring multi-source heterogeneous data in real time through an Internet of things sensing network deployed in a basic community, wherein the multi-source heterogeneous data comprises resident behavior tracks, public facility states and environment monitoring indexes, and performing abnormal noise rejection and space-time alignment processing on the multi-source heterogeneous data by adopting a self-adaptive space-time filtering algorithm to generate a standardized multi-mode sensing data stream; Based on the multi-mode sensing data flow, semantic ambiguity and logic conflict in the data are detected through conflict resolution knowledge graph, a data cleaning model is constructed by utilizing a dynamic rule engine, topology structure matching and duplication removal processing are carried out on repeated data in combination with a graph neural network, and a time-space associated cleaned data set is output; Inputting the cleaned data set into a multi-scale space-time encoder, extracting space-time correlation features in the data by fusing a time convolution network and a graph annotation force mechanism, and aligning evolution trends of different mode data based on a dynamic time warping algorithm to generate a feature tensor comprising region hot spot distribution and risk propagation paths; the process for generating the characteristic tensor comprising the regional hot spot distribution and the risk propagation path comprises the steps of inputting the cleaned data set into a multi-scale space-time coder, extracting a periodic mode of resident behavior tracks through a time convolution network, capturing a spatial dependency relationship among nodes of public facilities by using a graph attention mechanism to generate a primary characteristic vector, aligning an environment monitoring index with an evolution trend of the resident behavior tracks by using a dynamic time warping algorithm according to the primary characteristic vector, eliminating sensor acquisition delay through elastic matching of curvature constraint to generate a multi-mode characteristic matrix, carrying out Fourier transformation on the multi-mode characteristic matrix, extracting frequency domain energy distribution characteristics, further calculating a probability density function of the regional hot spot distribution, marking the region exceeding a preset threshold as a risk high-incidence region, simulating a risk propagation path along a facility network by using a graph diffusion model according to the risk high-incidence region, and further generating the risk propagation path; Based on the characteristic tensor, constructing a dynamic risk knowledge graph, carrying out probability prediction on potential risks in basic social treatment by using a causal reasoning model, simulating intervention effects of different treatment strategies by using a cross-modal countermeasure generation network, and outputting a decision matrix containing risk levels and optimal intervention paths; Inputting the decision matrix into a strategy optimization engine, synchronously optimizing treatment efficiency, resource consumption and resident satisfaction indexes by adopting a multi-target particle swarm algorithm, generating a preliminary treatment instruction set, virtually deducting the instruction execution effect through a digital twin platform, dynamically adjusting the instruction priority according to the deduction result, and obtaining a layered treatment instruction set to form an intelligent treatment strategy chain; Based on the real-time treatment feedback data after the layered treatment instruction set is executed, the conflict instructions in the strategy chain are automatically corrected through a fuzzy reinforcement learning algorithm, and strategy iteration tracks are recorded in a blockchain evidence storage network, so that verifiability and traceability of the treatment process are realized.
2. The method of claim 1, wherein the process of generating the multi-modal perceived data stream by anomaly noise rejection and spatio-temporal alignment processing of the multi-source heterogeneous data based on an adaptive spatio-temporal filtering algorithm comprises: Based on a sliding window mechanism in the self-adaptive space-time filtering algorithm, carrying out local variance analysis on the data flow of each sensor node, marking abnormal points exceeding a preset range as noise candidate areas, and generating an initial noise mask matrix; Performing space-time alignment on the environment monitoring index and the public facility state data, and introducing space-time density of resident movement tracks as a weight factor to correct an alignment path to generate a multi-mode data block; And inputting the multi-mode data block into a multi-channel self-encoder for characteristic weight distribution, so as to generate a multi-mode sensing data stream.
3. The method of claim 2, wherein the steps of detecting semantic ambiguity and logical conflict based on the multi-modal perceived data stream, constructing a data cleansing model using a dynamic rule engine, performing topology matching and deduplication on the duplicate data, and outputting the cleansed data set comprise: Extracting semantic triples according to entity relations in the multi-mode perception data stream, and constructing a conflict resolution knowledge graph by combining a domain knowledge base, wherein nodes in the conflict resolution knowledge graph represent entities or events, and edges represent logical dependency relations; According to the logic dependency relationship, detecting semantic ambiguity by using a dynamic rule engine, calculating conflict confidence, and generating a logic conflict event set; and inputting the logic conflict event set into a graph neural network, performing topology structure matching on repeated events, aggregating redundant nodes, and outputting a cleaned data set.
4. A method according to claim 3, wherein the process of constructing a dynamic risk knowledge-graph from the feature tensors and generating the decision matrix further comprises: Constructing a dynamic risk knowledge graph according to the risk propagation path in the feature tensor, calculating the conditional probability of each risk node by using a causal reasoning model, and generating a probability prediction result of a risk conducting chain; Inputting the probability prediction result into a cross-modal countermeasure generation network, and further outputting a strategy set with a weight score; And performing pareto front analysis on the strategy set to generate a decision matrix containing the optimal intervention path and risk level.
5. The method of claim 4, wherein forming the intelligent abatement strategy chain based on the decision matrix comprises: Inputting the decision matrix into a strategy optimization engine, further generating a preliminary treatment instruction set, mapping the preliminary treatment instruction set to a digital twin platform, and generating a deduction evaluation report; and calculating the emergency degree score of each instruction based on the fuzzy controller according to the deduction evaluation report, dynamically adjusting the instruction priority, generating a layered treatment instruction set, and further forming an intelligent treatment strategy chain.
6. The method of claim 5, wherein implementing verifiability and traceability of the abatement process based on the layered abatement instruction set comprises: according to the real-time treatment feedback data after the layered treatment instruction set is executed, detecting conflict instructions in an intelligent treatment strategy chain through a fuzzy Q-learning algorithm, calculating conflict intensity values among the instructions, and generating a conflict instruction set; inputting the conflict instruction set into a depth deterministic strategy gradient model, and further outputting an optimized intelligent treatment strategy; And packaging the strategy iteration track into an intelligent contract, further generating a non-tamperable record with a time stamp and a hash value, and realizing the verifiability and traceability of the treatment process.
7. A digital governance system for a substrate society, said system comprising: The acquisition module is used for acquiring multi-source heterogeneous data in real time through an Internet of things sensing network deployed in a basic community, wherein the multi-source heterogeneous data comprises resident behavior tracks, public facility states and environment monitoring indexes, abnormal noise rejection and space-time alignment processing are carried out on the multi-source heterogeneous data, and a multi-mode sensing data stream is generated; the detection module is used for detecting the multi-mode sensing data flow, constructing a data cleaning model, carrying out topological structure matching and de-duplication processing on repeated data by combining a graph neural network, and outputting a cleaned data set; The extraction module is used for extracting space-time correlation features in the cleaned data set, aligning evolution trends of different mode data based on a dynamic time warping algorithm and generating feature tensors; the process for generating the characteristic tensor comprising the regional hot spot distribution and the risk propagation path comprises the steps of inputting the cleaned data set into a multi-scale space-time coder, extracting a periodic mode of resident behavior tracks through a time convolution network, capturing a spatial dependency relationship among nodes of public facilities by using a graph attention mechanism to generate a primary characteristic vector, aligning an environment monitoring index with an evolution trend of the resident behavior tracks by using a dynamic time warping algorithm according to the primary characteristic vector, eliminating sensor acquisition delay through elastic matching of curvature constraint to generate a multi-mode characteristic matrix, carrying out Fourier transformation on the multi-mode characteristic matrix, extracting frequency domain energy distribution characteristics, further calculating a probability density function of the regional hot spot distribution, marking the region exceeding a preset threshold as a risk high-incidence region, simulating a risk propagation path along a facility network by using a graph diffusion model according to the risk high-incidence region, and further generating the risk propagation path; the prediction module is used for constructing a dynamic risk knowledge graph, carrying out probability prediction on potential risks in basic social treatment, and further generating a decision matrix; The generation module is used for inputting the decision matrix into the strategy optimization engine, generating a preliminary treatment instruction set, virtually deducting the instruction execution effect through the digital twin platform, dynamically adjusting the instruction priority according to the deduction result, obtaining a layered treatment instruction set, and further forming an intelligent treatment strategy chain; And the correction module is used for realizing verifiability and traceability of the treatment process according to the layered treatment instruction set.
8. A storage medium having a computer program stored therein, wherein the computer program is arranged to perform the method of any of claims 1-6 when run.
9. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of claims 1-6.

Description

Digital treatment method and system for basic society Technical Field The invention belongs to the technical field of digitization, in particular to a base society digitization treatment method and system. Background With the rapid development of information technology, digital governance is becoming an important means for improving public management efficiency and service quality. Basic social management is taken as a basic link of social management, relates to a plurality of aspects of resident life, public facility management, environmental protection and the like, and faces a plurality of challenges such as information island, uneven resource allocation, lagged management means and the like. The traditional basic management mode often depends on manual operation and experience judgment, is difficult to effectively cope with complex and changeable social environment and resident demands, and further causes low management efficiency, resource waste and resident satisfaction reduction. Under the background, the rise of the Internet of things and big data technology provides a new opportunity for the basic-level society management. Through deploying the sensor of the internet of things, multi-source heterogeneous data such as resident behaviors, public facility states and environmental indexes can be collected in real time, and support is provided for scientific decision making and dynamic treatment. However, in practice, there are often noise, ambiguity, and logic conflicts in these data, and how to effectively clean and process these data, and extracting useful information to guide governance decisions, is a key to achieving digital governance. Disclosure of Invention The invention aims to provide a digital treatment method and system for a basic society, which are used for solving the defects in the prior art, and can realize a digital treatment scheme with high efficiency, intelligence and traceability so as to improve the efficiency and service quality of the social treatment. One embodiment of the application provides a digital treatment method for a base society, which comprises the following steps: Acquiring multi-source heterogeneous data in real time through an Internet of things sensing network deployed in a basic community, wherein the multi-source heterogeneous data comprises resident behavior tracks, public facility states and environment monitoring indexes, and performing abnormal noise rejection and space-time alignment processing on the multi-source heterogeneous data by adopting a self-adaptive space-time filtering algorithm to generate a standardized multi-mode sensing data stream; Based on the multi-mode sensing data flow, semantic ambiguity and logic conflict in the data are detected through conflict resolution knowledge graph, a data cleaning model is constructed by utilizing a dynamic rule engine, topology structure matching and duplication removal processing are carried out on repeated data in combination with a graph neural network, and a time-space associated cleaned data set is output; Inputting the cleaned data set into a multi-scale space-time encoder, extracting space-time correlation features in the data by fusing a time convolution network and a graph annotation force mechanism, and aligning evolution trends of different mode data based on a dynamic time warping algorithm to generate a feature tensor comprising region hot spot distribution and risk propagation paths; Based on the characteristic tensor, constructing a dynamic risk knowledge graph, carrying out probability prediction on potential risks in basic social treatment by using a causal reasoning model, simulating intervention effects of different treatment strategies by using a cross-modal countermeasure generation network, and outputting a decision matrix containing risk levels and optimal intervention paths; Inputting the decision matrix into a strategy optimization engine, synchronously optimizing treatment efficiency, resource consumption and resident satisfaction indexes by adopting a multi-target particle swarm algorithm, generating a preliminary treatment instruction set, virtually deducting the instruction execution effect through a digital twin platform, dynamically adjusting the instruction priority according to the deduction result, and obtaining a layered treatment instruction set to form an intelligent treatment strategy chain; Based on the real-time treatment feedback data after the layered treatment instruction set is executed, the conflict instructions in the strategy chain are automatically corrected through a fuzzy reinforcement learning algorithm, and strategy iteration tracks are recorded in a blockchain evidence storage network, so that verifiability and traceability of the treatment process are realized. According to one preferred embodiment of the present invention, the process of generating the multi-modal sensing data stream by performing abnormal noise rejection and space-time alignment proce