An intelligent management system and method for evidence contamination platforms

Abstract

This invention discloses an intelligent management system and method for an evidence contamination platform, relating to the field of intelligent management technology. The invention collects historical data from the platform, preprocesses it, and then constructs an evidence value assessment model and a pollution risk assessment model based on the historical data. The evidence value assessment model integrates the PageRank algorithm and a weighted linear regression algorithm to generate a comprehensive evidence value score. The pollution risk assessment model combines a graph neural network algorithm and a random forest algorithm to output a comprehensive pollution risk score. The two scores are mapped to a two-dimensional coordinate system, with evidence value as the horizontal axis and pollution risk as the vertical axis, dividing the system into four quadrants: high value-high risk, high value-low risk, low value-high risk, and low value-low risk. This forms a dynamic dual-coordinate decision-making mechanism for evidence value and pollution risk, allowing for different processing methods.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology, specifically an intelligent management system and method applied to an evidence contamination platform. Background Technology

[0002] In many fields today, the correctness, reliability, and value assessment of evidence are crucial. With the rapid development of information technology, the forms of evidence are becoming increasingly diverse, and the environments in which it is generated, disseminated, and used are becoming more complex. This has led to a growing problem of evidence contamination, posing a serious challenge to evidence-based decision-making. How to effectively manage evidence contamination and ensure the integrity and reliability of the chain of evidence has become a critical issue that urgently needs to be addressed.

[0003] Traditional evidence valuation techniques often rely on human experience or simple indicators, making them highly subjective and unable to quantify the true impact of evidence in complex network environments and its propagation effect in citation relationships. This fails to comprehensively reflect the actual value of the evidence. Pollution risk assessments often rely on a few isolated features for simple judgments, lacking a systematic analysis of the propagation paths, scope of impact, and potential hazards of pollution in complex interconnected networks. This assessment method not only fails to accurately characterize risk but also lacks quantitative means, resulting in qualitative or rough risk assessments rather than precise quantitative scores. Existing technologies typically lack a systematic framework that integrates evidentiary value and risk information, making it difficult to formulate differentiated intelligent management strategies. Evidence processing procedures are often uniform or categorized according to a single dimension, failing to fully consider the coupling relationship between value and risk. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent management system and method for evidence contamination platforms to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides an intelligent management method for an evidence contamination platform, comprising:

[0007] Historical data from the collection platform, including identified pollution events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals, are preprocessed.

[0008] Based on preprocessed historical data, an evidence value assessment model and a pollution risk assessment model are constructed. The evidence value assessment model uses a weighted scoring fusion model algorithm, which analyzes influence ranking and citation relationships using the PageRank algorithm to identify high-impact evidence and extract node importance. It also uses a weighted linear regression algorithm to analyze the number of citations by top journals and the frequency of evidence use. The comprehensive evidence value score is calculated by combining the PageRank score and the linear regression weight score. The pollution risk assessment model constructs a graph data structure, uses the Graph Neural Network (GNN) algorithm to extract risk features, and uses the Random Forest algorithm to build a classification model for non-graph features. The comprehensive pollution risk score is generated by fusing the GNN output and the classification model results.

[0009] The evidence value score and pollution risk score are mapped to a two-dimensional coordinate system, with the evidence value score as the horizontal axis and the pollution risk score as the vertical axis, to establish a dual-coordinate decision-making mechanism of evidence value and pollution risk. Different risk areas and thresholds are set, and the quadrants are divided into high-value-high-risk quadrant, high-value-low-risk quadrant, low-value-high-risk quadrant, and low-value-low-risk quadrant according to the update priority. Different processing strategies are assigned according to different quadrants.

[0010] In conjunction with the first aspect, in the first implementation of the first aspect of this application, the historical data collected by the platform includes identified pollution events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals. The data is preprocessed, including:

[0011] Collect historical data from the evidence contamination platform, including identified contamination events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals;

[0012] The collected historical data from the platform is preprocessed. Based on the data type and distribution, missing values ​​are filled by the mean or mode. Outlier data points are identified and processed using the Z-score statistical method. Duplicate records are deleted. Regular expressions are used to standardize and denoise the text data.

[0013] In conjunction with the first aspect, in the second implementation of the first aspect of this application, the evidence value assessment model uses a weighted scoring fusion model algorithm. It analyzes influence ranking and citation relationships using the PageRank algorithm to identify high-impact evidence and extract node importance. It also analyzes the number of times evidence is cited by top journals and the frequency of evidence use using a weighted linear regression algorithm, including:

[0014] The PageRank algorithm is used to iteratively calculate the evidence nodes and their citation relationship network. Based on the quality and quantity of nodes entering the chain, the relative influence of each evidence node in the network is evaluated and quantified, generating a node importance score based on the network structure. The calculation formula is as follows:

[0015] ;

[0016] Where PR(A) is the PageRank score of evidence node A, representing the influence of evidence node A in the evidence network; d is the damping coefficient, representing the probability that the user will continue to click the link; 1-d is the complement of the damping coefficient, representing the probability that the user will not continue to click the link; B(A) is the set of all evidence nodes pointing to evidence node A; PR(T) is the PageRank score of evidence node T, representing the influence of evidence node T in the evidence network; C(T) is the number of backlinks of evidence node T.

[0017] Collect and organize data on the number of times each piece of evidence is cited in top journals and its usage frequency within the platform, using the following formula:

[0018] ;

[0019] Where Y is the evidence value score, X1 is the number of times cited by top journals, X2 is the frequency of evidence use, β0 is the intercept, and β1 and β2 are regression coefficients, representing the degree of influence of each independent variable on the dependent variable. It is an error term;

[0020] A weighted linear regression model is constructed, with evidential value as the dependent variable and the number of times the evidence is cited in top-tier journals and the frequency of use of the evidence as independent variables. Different weights are assigned based on the historical revision effect or influence ranking of the evidence. A regression equation is fitted, these two factors are quantified, and an evidential value score based on statistical relationships is generated. The formula is as follows:

[0021] ;

[0022] w i Y is the weight of the i-th sample, set according to the historical correction effect or influence ranking of the evidence, adjusting the importance of this sample in the regression model; i It is the actual value of the i-th sample; is the predicted value of the i-th sample; n is the number of samples; this formula finds the regression coefficients and fits the regression equation by minimizing the weighted sum of squared residuals; where the weighted linear regression score LR is equal to the predicted value of the regression model. .

[0023] In conjunction with the first aspect, in the third implementation of the first aspect of this application, the calculation of the comprehensive evidentiary value score by combining the PageRank score and the linear regression weight score includes:

[0024] The PageRank score PR(A) and the weighted linear regression score LR are weighted and fused according to preset weights to generate the final comprehensive evidence value score V, as shown in the formula:

[0025] ;

[0026] Where V is the final generated comprehensive evidence value score, representing the overall value of a certain evidence node; α is a weighting coefficient between 0 and 1, representing the proportion of the PageRank score PR(A) in the final score V; 1-α represents the proportion of the weighted linear regression score LR.

[0027] In conjunction with the first aspect, in the fourth implementation of the first aspect of this application, the pollution risk assessment model constructs a graph data structure, uses a graph neural network algorithm (GNN) to extract risk features, and uses a random forest algorithm to establish a classification model for non-graph features, including:

[0028] Evidence entities, users, and interaction relationships are abstracted into nodes and edges in a graph, forming graph data that can represent the flow and correlation of evidence. The constructed graph data is input into a graph neural network model. The GNN learns the representation of nodes in the graph structure layer by layer by aggregating the features of the nodes themselves and the features of the neighboring nodes, thereby capturing risk patterns.

[0029] Collect and process the non-graph structure features of the evidence, and train a random forest classifier; the random forest determines whether the evidence has a risk of contamination by constructing multiple decision trees and combining the prediction results.

[0030] In conjunction with the first aspect, in the fifth implementation of the first aspect of this application, the step of fusing the GNN output and the classification model results to generate a comprehensive pollution risk score includes:

[0031] The structured risk features extracted by GNN are fused with the non-graph features utilized by random forest to comprehensively assess the contamination risk or probability of evidence. The formula is as follows:

[0032] ;

[0033] Wherein, FRS is the final integrated pollution risk score; GS is the pollution risk score output by the graph neural network model; RS is the pollution risk score output by the non-graph feature model; θ is a weighting coefficient used to adjust the relative importance of GS and RS, with a value ranging from 0 to 1.

[0034] In conjunction with the first aspect, in the sixth implementation of the first aspect of this application, the mapping of the evidence value score and the pollution risk score to a two-dimensional coordinate system, with the evidence value score as the horizontal axis and the pollution risk score as the vertical axis, to establish an evidence value-pollution risk dual-coordinate decision-making mechanism, includes:

[0035] The evidence value score and pollution risk score are mapped onto a two-dimensional coordinate system. The horizontal axis X represents the evidence value score, reflecting the importance, relevance, and probative value of the evidence to the case. The higher the score, the greater the potential role of the evidence in the case. The vertical axis Y represents the pollution risk level score, reflecting the possibility or severity of the evidence being contaminated. The higher the score, the lower the reliability of the evidence and the greater the risk of misleading the judgment of the case.

[0036] By plotting these two scores for each piece of evidence onto the coordinate system, an evidence point is formed. The specific location of the evidence point in the coordinate system reflects the overall value and risk of the evidence.

[0037] In conjunction with the first aspect, in the seventh implementation of the first aspect of this application, the setting of different risk zones and thresholds, divided into high-value-high-risk quadrants, high-value-low-risk quadrants, low-value-high-risk quadrants, and low-value-low-risk quadrants according to update priority, includes:

[0038] Based on the established two-dimensional coordinate system of evidence value and pollution risk, thresholds are set for evidence value score and pollution risk score, and the four quadrants are set as high value-high risk quadrant, low value-high risk quadrant, low value-low risk quadrant and high value-low risk quadrant respectively.

[0039] The high-value-high-risk quadrant is located in the upper right corner of the coordinate system. Evidence with both evidentiary value and contamination risk exceeds its respective preset threshold. This type of evidence has a significant impact on the case, but its reliability is questionable, and it requires the highest priority for attention and processing. The high-value-low-risk quadrant is located in the lower right corner of the coordinate system. Evidence with evidentiary value exceeds the preset evidentiary value threshold, while its contamination risk is below the preset contamination risk threshold. This type of evidence has a significant impact on the case but a low contamination risk, and requires the second priority for identification and processing. The low-value-high-risk quadrant is located in the upper left corner of the coordinate system. Evidence with evidentiary value is below the preset evidentiary value threshold, while its contamination risk is above the preset contamination risk threshold. This type of evidence has little interference but a high risk profile, and requires the third priority for identification and processing. The low-value-low-risk quadrant is located in the lower left corner of the coordinate system. Evidence with both evidentiary value and contamination risk are below their respective preset thresholds, and this type of evidence has the lowest priority for updating.

[0040] In conjunction with the first aspect, in the eighth implementation of the first aspect of this application, the allocation of different processing strategies according to different quadrants includes:

[0041] For evidence located in the high-value-high-risk quadrant, when the frequency of use of the evidence exceeds a preset frequency threshold, the highest level of handling measures are initiated; comprehensive pollution source tracing and characteristic analysis are performed, and this information is compared with the platform's historical pollution event database to identify pollution sources and potential transmission patterns; cross-disciplinary experts are organized to assess the pollution impact, and the scope and depth of pollution are determined by referring to the correction effect records of similar pollution to low-value-high-risk evidence; mandatory pollution elimination or evidence isolation measures are implemented, and the entire process is recorded;

[0042] For evidence located in the high-value-low-risk quadrant, when the frequency of use of such evidence exceeds a preset frequency threshold, intermediate processing measures are initiated; its low-risk status is confirmed, and it is prioritized for inclusion in case analysis, with its utilization and correction effects recorded; it is prioritized for presentation as key evidence, with its specific role and effect in case progression recorded; routine monitoring is implemented, and the monitoring data is compared and analyzed with the monitoring data of high-value-high-risk evidence to form a risk warning linkage; it is assessed whether additional protection measures are needed to prevent future contamination, and the processing procedures and effects are recorded and fed back to the high-value-high-risk quadrant for reference;

[0043] For evidence located in the low-value-high-risk quadrant, when the frequency of use of the evidence exceeds a preset frequency threshold, basic processing measures are initiated; risk technology assessment is conducted to confirm the pollution status and high-risk attributes; pollution elimination is performed or the evidence is marked as unacceptable and excluded, with removal being given priority to reduce risk; the processing results are archived and fed back to the model update, while the processing strategy is used as a reference and linked to the decision-making process of the low-value-low-risk quadrant.

[0044] For evidence located in the low-value-low-risk quadrant, when the frequency of use of the evidence is lower than a preset frequency threshold, minimum processing measures are initiated; after confirming low risk, a standardized archiving process is adopted, a low-value tag is added, and the evidence is retrieved under specific search requests to monitor its usage. The processing flow is then fed back to the high-value-low-risk quadrant.

[0045] Secondly, the present invention provides an intelligent management system applied to an evidence contamination platform, comprising:

[0046] Data Acquisition and Preprocessing Module: Includes a data acquisition unit and a data preprocessing unit; wherein, the data acquisition unit collects historical data from the platform; the data preprocessing unit preprocesses the collected historical data from the platform, identifies and processes missing values, detects and processes abnormal data points, deletes duplicate records, and standardizes and denoises the text data;

[0047] The evidence value assessment module includes an influence analysis unit, a statistical relationship analysis unit, and an evidence value score fusion unit. The influence analysis unit uses the PageRank algorithm to analyze evidence nodes and citation relationship networks, calculating the PageRank score for each evidence node. The statistical relationship analysis unit collects and organizes data on the number of times evidence is cited by top journals and its usage frequency within the platform, constructs a weighted linear regression model, analyzes the impact of these two factors on the evidence value, and generates an evidence value score. The evidence value score fusion unit weights and fuses the PageRank score and the weighted linear regression score according to preset weights to generate the final comprehensive evidence value score.

[0048] The pollution risk assessment module includes a structured risk extraction unit, a non-graph feature processing unit, and a risk score fusion unit. The structured risk extraction unit abstracts evidence entities, users, and their interactions into nodes and edges in a graph data structure, forming graph data. This graph data is then input into a graph neural network (GNN) model to capture and output a risk feature score based on the graph structure. The non-graph feature processing unit collects and processes the non-graph structural features of the evidence, trains a random forest classifier, determines whether the evidence poses a pollution risk, and outputs a risk score based on non-graph features. The risk score fusion unit merges the structured risk score with the non-graph feature risk score to generate the final comprehensive pollution risk score.

[0049] The evidence value-pollution risk coordinate system construction module includes a coordinate mapping unit and a quadrant division unit. The coordinate mapping unit receives the evidence value score and the pollution risk score, and maps each evidence point to a two-dimensional coordinate system, where the horizontal axis represents the evidence value score and the vertical axis represents the pollution risk score. The quadrant division unit divides the two-dimensional coordinate system into four quadrants according to a preset threshold: high value-high risk quadrant, high value-low risk quadrant, low value-high risk quadrant, and low value-low risk quadrant.

[0050] The evidence processing strategy decision-making module includes a priority and hierarchy determination unit, a threshold judgment and status confirmation unit, and a strategy matching and generation unit. The priority and hierarchy determination unit determines the processing priority order and level of evidence based on its quadrant in the coordinate system, according to preset priority rules and processing levels. The threshold judgment and status confirmation unit checks whether the specific attributes of evidence in a specific quadrant meet the prescribed threshold conditions, and for evidence that meets the conditions, confirms its specific status within that quadrant. The strategy matching and generation unit combines the evidence's quadrant, the determined processing level, and the results of the threshold judgment and status confirmation unit to accurately match and generate corresponding processing strategies from a predefined strategy library, generating specific, executable operations or suggestions, and implementing feedback and linkage between strategies.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] 1. This invention achieves a comprehensive quantitative score of evidence value by combining the PageRank algorithm and a weighted linear regression model; it uses graph neural networks to process structured risk features and combines random forests to process non-graph features, and merges the results of the two to generate a comprehensive pollution risk score.

[0053] 2. This invention constructs a dual-coordinate decision-making mechanism of evidence value and pollution risk by mapping the quantified evidence value score and pollution risk score to a two-dimensional coordinate system, and divides the evidence into four quadrants according to a preset threshold.

[0054] 3. This invention designs clear priorities and differentiated processing strategies for evidence in different quadrants, taking into account the current state and dynamic conditions of the evidence. The processing flow and results of different quadrants provide feedback to each other, forming a linkage and closed-loop optimization between strategies. Attached Figure Description

[0055] Figure 1 This is a schematic diagram illustrating the steps of an intelligent management method for an evidence contamination platform according to the present invention;

[0056] Figure 2 This is a system architecture diagram of an intelligent management system applied to an evidence contamination platform according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Example: Figures 1-2 As shown, the present invention provides a technical solution.

[0059] like Figure 1 A schematic diagram illustrating the steps of an intelligent management method applied to an evidence contamination platform illustrates that this invention provides an intelligent management method applied to an evidence contamination platform, comprising:

[0060] Step S100: Collect historical data from the platform, including identified pollution events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals, and preprocess the data;

[0061] Specifically, collect historical data from the evidence contamination platform, including identified contamination events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals;

[0062] The collected historical data from the platform is preprocessed. Based on the data type and distribution, missing values ​​are filled by the mean or mode. Outlier data points are identified and processed using the Z-score statistical method. Duplicate records are deleted. Regular expressions are used to standardize and denoise the text data.

[0063] In one specific embodiment, raw logs and event records from January 2020 to December 2024 were exported from the backend database of the evidence contamination platform. A total of 237 identified contamination event records were collected. Usage frequency data for 15,842 evidence items on the platform were also collected. It was found that evidence E-7821 was used 187 times throughout 2024, while evidence E-9534 was only used 3 times. Contamination event ID-105 was marked as "malicious script injection, targeting evidence E-7821." Records show that evidence E-7821 was cited by 3 other pieces of evidence, causing the contamination to spread to these 3 cited pieces of evidence. Correction effect data includes correction records for contamination event ID-105, showing that the accuracy of evidence E-7821 improved from 72% to 95% after correction. Influence ranking data comes from the platform's internal evaluation model, with evidence E-7821 ranking 12th in influence. Regarding citations by top journals, evidence E-7821 was cited 8 times. During the preprocessing stage, it was discovered that the pollution source characteristic field of evidence E-4567 was missing. Since this field is categorical data, the mode "error label" was used to fill it in. When processing the usage frequency data using the Z-score method, it was found that the usage frequency of evidence E-7821 (187 times) far exceeded the mean of 45 times and had a Z-score greater than 3, and was identified as an outlier. After verification, it was confirmed that its high-frequency usage was related to the importance of the evidence, so this data point was retained. However, other obviously erroneous records with an absolute Z-score greater than 3 were corrected or deleted. Three completely duplicate evidence records caused by system errors were deleted, and regular expressions were used to clean up and standardize redundant spaces and special characters in the text fields.

[0064] Step S200: Based on the preprocessed historical data, construct an evidence value assessment model and a pollution risk assessment model; the evidence value assessment model uses a weighted scoring fusion model algorithm, analyzes the influence ranking and citation relationship through the PageRank algorithm, identifies high-impact evidence, extracts node importance, and analyzes the number of times cited by top journals and the frequency of evidence use through a weighted linear regression algorithm; combining the PageRank score and the linear regression weight score, calculate the comprehensive evidence value score;

[0065] Specifically, the PageRank algorithm is used to iteratively calculate the evidence nodes and their citation relationship network. Based on the quality and quantity of nodes entering the chain, the relative influence of each evidence node in the network is evaluated and quantified, generating a node importance score based on the network structure. The calculation formula is as follows:

[0066] ;

[0067] Where PR(A) is the PageRank score of evidence node A, representing the influence of evidence node A in the evidence network; d is the damping coefficient, representing the probability that the user will continue to click the link; 1-d is the complement of the damping coefficient, representing the probability that the user will not continue to click the link; B(A) is the set of all evidence nodes pointing to evidence node A; PR(T) is the PageRank score of evidence node T, representing the influence of evidence node T in the evidence network; C(T) is the number of backlinks of evidence node T.

[0068] Collect and organize data on the number of times each piece of evidence is cited in top journals and its usage frequency within the platform, using the following formula:

[0069] ;

[0070] Where Y is the evidence value score, X1 is the number of times cited by top journals, X2 is the frequency of evidence use, β0 is the intercept, and β1 and β2 are regression coefficients, representing the degree of influence of each independent variable on the dependent variable. It is an error term;

[0071] A weighted linear regression model is constructed, with evidential value as the dependent variable and the number of times the evidence is cited in top-tier journals and the frequency of use of the evidence as independent variables. Different weights are assigned based on the historical revision effect or influence ranking of the evidence. A regression equation is fitted, these two factors are quantified, and an evidential value score based on statistical relationships is generated. The formula is as follows:

[0072] ;

[0073] w i Y is the weight of the i-th sample, set according to the historical correction effect or influence ranking of the evidence, adjusting the importance of this sample in the regression model; i It is the actual value of the i-th sample; is the predicted value of the i-th sample; n is the number of samples; this formula finds the regression coefficients and fits the regression equation by minimizing the weighted sum of squared residuals; where the weighted linear regression score LR is equal to the predicted value of the regression model. .

[0074] Furthermore, the PageRank score PR(A) and the weighted linear regression score LR are weighted and fused according to preset weights to generate the final comprehensive evidence value score V, as shown in the formula:

[0075] ;

[0076] Where V is the final generated comprehensive evidence value score, representing the overall value of a certain evidence node; α is a weighting coefficient between 0 and 1, representing the proportion of the PageRank score PR(A) in the final score V; 1-α represents the proportion of the weighted linear regression score LR.

[0077] In one specific embodiment, two models are constructed using preprocessed historical data. For the evidence value assessment model, the PageRank algorithm is applied to analyze the citation relationship network between evidence. Evidence E-7821 has a high number and quality of incoming links in the network. After multiple iterations, its PageRank score PR(A) finally converged to 0.087. Simultaneously, it was found that this evidence was cited 8 times by top journals (X1) and used 187 times within the platform during the statistical period. A weighted linear regression model is constructed. Based on the significant improvement in accuracy after correction and its 12th place ranking in influence, evidence E-7821 is assigned a higher weight w. i The model fit yielded a regression equation of Y = 0.35X1 + 0.12X2 + 0.5, where β0 = 0.5, β1 = 0.35, and β2 = 0.12. Substituting X1 = 8 and X2 = 187, the weighted linear regression prediction score Ŷi for this evidence was 25.74. The PageRank score PR(A) = 0.087 and the weighted linear regression score LR = 25.74 were then weighted and fused. The preset weight coefficient α was set to 0.3, indicating that the network structure influence accounted for 30%, so 1 - α = 0.7. The comprehensive evidence value score V for evidence E-7821 was approximately 18.044. Through similar calculations, a comprehensive value score V could be calculated for all evidence within the platform.

[0078] Step S300: Construct a graph data structure for the pollution risk assessment model, use the graph neural network algorithm (GNN) to extract risk features, and use the random forest algorithm to build a classification model for non-graph features; fuse the GNN output and the classification model results to generate a comprehensive pollution risk score.

[0079] Specifically, evidence entities, users, and interaction relationships are abstracted into nodes and edges in a graph to form graph data that can represent the flow and correlation of evidence. The constructed graph data is then input into a graph neural network model. The GNN learns the representation of nodes in the graph structure layer by layer by aggregating the features of the nodes themselves and the features of the neighboring nodes, thereby capturing risk patterns.

[0080] Collect and process the non-graph structure features of the evidence, and train a random forest classifier; the random forest determines whether the evidence has a risk of contamination by constructing multiple decision trees and combining the prediction results.

[0081] Furthermore, the structured risk features extracted by GNN are fused with the non-graph features utilized by random forest to comprehensively assess the contamination risk or probability of the evidence. The formula is as follows:

[0082] ;

[0083] Wherein, FRS is the final integrated pollution risk score; GS is the pollution risk score output by the graph neural network model; RS is the pollution risk score output by the non-graph feature model; θ is a weighting coefficient used to adjust the relative importance of GS and RS, with a value ranging from 0 to 1.

[0084] In one specific embodiment, the evidence items, users, and their interactions within the evidence contamination platform are constructed as a graph data structure. Evidence E-7821 is represented as a node, with features including type, creation time, and source. User U-1024 downloaded this evidence, establishing an edge between the user U-1024 node and the evidence E-7821 node, labeling the interaction type as "download." Graph data containing approximately 15,842 evidence nodes, 10,000 user nodes, and approximately 200,000 interaction edges between them is input into a graph neural network (GNN) model. This GNN model is trained with three layers, learning and outputting a risk representation for each node by aggregating its own features and those of its neighboring nodes. For evidence E-7821, the risk score (GS) output by the GNN model is 0.72. Simultaneously, non-graph structure features of evidence E-7821 are collected, showing a usage frequency of 187 times, 8 citations by top-tier journals, and a PageRank score of 0.087. A random forest classifier is trained using these features, and the model outputs a probability or score (RS) representing pollution risk. The random forest model predicts a pollution risk score (RS) of 0.65 for evidence E-7821. The structured risk feature score (GS) of the GNN is then fused with the non-graph feature score (RS) of the random forest. A weighting coefficient θ is set to 0.6 to emphasize graph structure information. The comprehensive pollution risk score (FRS) for evidence E-7821 is calculated to be 0.692 according to the formula. Using this method, a comprehensive pollution risk score (FRS) can be calculated for all evidence within the platform.

[0085] Step S400: Map the evidence value score and pollution risk score to a two-dimensional coordinate system, with the evidence value score as the horizontal axis and the pollution risk score as the vertical axis, to establish an evidence value-pollution risk dual-coordinate decision-making mechanism, set different risk areas and thresholds, and divide them into high value-high risk quadrant, high value-low risk quadrant, low value-high risk quadrant and low value-low risk quadrant according to the update priority order.

[0086] Specifically, the evidence value score and pollution risk score are mapped onto a two-dimensional coordinate system. The horizontal axis X represents the evidence value score, reflecting the importance, relevance, and probative value of the evidence to the case; the higher the score, the greater the potential role of the evidence in the case. The vertical axis Y represents the pollution risk level score, reflecting the possibility or severity of the evidence being contaminated; the higher the score, the lower the reliability of the evidence and the greater the risk of misleading the judgment of the case.

[0087] By plotting these two scores for each piece of evidence onto the coordinate system, an evidence point is formed. The specific location of the evidence point in the coordinate system reflects the overall value and risk of the evidence.

[0088] Furthermore, based on the established two-dimensional coordinate system of evidence value and pollution risk, thresholds are set for evidence value scores and pollution risk scores, and the four quadrants are respectively set as high value-high risk quadrant, low value-high risk quadrant, low value-low risk quadrant and high value-low risk quadrant.

[0089] The high-value-high-risk quadrant is located in the upper right corner of the coordinate system. Evidence with both evidentiary value and contamination risk exceeds its respective preset threshold. This type of evidence has a significant impact on the case, but its reliability is questionable, and it requires the highest priority for attention and processing. The high-value-low-risk quadrant is located in the lower right corner of the coordinate system. Evidence with evidentiary value exceeds the preset evidentiary value threshold, while its contamination risk is below the preset contamination risk threshold. This type of evidence has a significant impact on the case but a low contamination risk, and requires the second priority for identification and processing. The low-value-high-risk quadrant is located in the upper left corner of the coordinate system. Evidence with evidentiary value is below the preset evidentiary value threshold, while its contamination risk is above the preset contamination risk threshold. This type of evidence has little interference but a high risk profile, and requires the third priority for identification and processing. The low-value-low-risk quadrant is located in the lower left corner of the coordinate system. Evidence with both evidentiary value and contamination risk are below their respective preset thresholds, and this type of evidence has the lowest priority for updating.

[0090] In one specific embodiment, the threshold for the evidence value score V is set to 20, and the threshold for the pollution risk score FRS is set to 0.6. The comprehensive evidence value score V = 18.044 and the comprehensive pollution risk score FRS = 0.692 of evidence E-7821 are mapped onto a two-dimensional coordinate system. Since V = 18.0441 is below the threshold of 20, and FRS = 0.692 is above the threshold of 0.6, the point for evidence E-7821 is plotted in the upper left corner of the coordinate system. According to the preset quadrant division rules, this area corresponds to the low-value-high-risk quadrant. This means that although evidence E-7821 has relatively low value to the case, its pollution risk is high. Based on the priority setting of this quadrant, evidence E-7821 is classified as evidence requiring third-priority identification and processing. Evidence E-1234, with an evidence value score of V=25.5 and a pollution risk score of FRS=0.55, will fall in the high value-low risk quadrant, while evidence E-5678, with an evidence value score of V=15.2 and a pollution risk score of FRS=0.7, will fall in the low value-high risk quadrant. They will be assigned different processing priorities.

[0091] In this way, all evidence on the platform is located in a coordinate system based on its evidence value score and pollution risk score, and is clearly divided into four different quadrants, laying the foundation for subsequent intelligent management strategy allocation.

[0092] Step S500: Assign different processing strategies according to different quadrants.

[0093] Specifically, for evidence located in the high-value-high-risk quadrant, when the frequency of use of the evidence exceeds a preset frequency threshold, the highest level of handling measures are initiated; comprehensive pollution source tracing and characteristic analysis are performed, and this information is compared with the platform's historical pollution event database to identify pollution sources and potential transmission patterns; cross-disciplinary experts are organized to assess the pollution impact, and the scope and depth of pollution are determined by referring to the correction effect records of similar pollution to low-value-high-risk evidence; mandatory pollution elimination or evidence isolation measures are implemented, and the entire process is recorded.

[0094] For evidence located in the high-value-low-risk quadrant, when the frequency of use of such evidence exceeds a preset frequency threshold, intermediate processing measures are initiated; its low-risk status is confirmed, and it is prioritized for inclusion in case analysis, with its utilization and correction effects recorded; it is prioritized for presentation as key evidence, with its specific role and effect in case progression recorded; routine monitoring is implemented, and the monitoring data is compared and analyzed with the monitoring data of high-value-high-risk evidence to form a risk warning linkage; it is assessed whether additional protection measures are needed to prevent future contamination, and the processing procedures and effects are recorded and fed back to the high-value-high-risk quadrant for reference;

[0095] For evidence located in the low-value-high-risk quadrant, when the frequency of use of the evidence exceeds a preset frequency threshold, basic processing measures are initiated; risk technology assessment is conducted to confirm the pollution status and high-risk attributes; pollution elimination is performed or the evidence is marked as unacceptable and excluded, with removal being given priority to reduce risk; the processing results are archived and fed back to the model update, while the processing strategy is used as a reference and linked to the decision-making process of the low-value-low-risk quadrant.

[0096] For evidence located in the low-value-low-risk quadrant, when the frequency of use of the evidence is lower than a preset frequency threshold, minimum processing measures are initiated; after confirming low risk, a standardized archiving process is adopted, a low-value tag is added, and the evidence is retrieved under specific search requests to monitor its usage. The processing flow is then fed back to the high-value-low-risk quadrant.

[0097] In one specific embodiment, evidence E-7821 has a value score (V) of 18.0441, which is below the threshold of 20, and a risk score (FRS) of 0.692, which is above the threshold of 0.6, thus placing it in the low-value-high-risk quadrant. An examination of its attributes revealed a usage frequency of 210 times, exceeding the preset usage frequency threshold of 150 times. According to the strategy, basic processing measures were initiated, including risk technical assessment, confirming that the evidence originated from a known high-risk pollution source and showed signs of data tampering, thus confirming its high-risk attribute. Pollution elimination measures were implemented, removing the relevant pollution data fragments and marking the evidence as "corrected, but for reference only." The processing results were archived and fed back to the model update database, while the processing strategy was recorded as a reference for future decision-making processes involving similar low-value-low-risk evidence.

[0098] Evidence E-1234, with a V=25.5 (above the threshold of 20) and FRS=0.55 (below the threshold of 0.6), falls within the high-value-low-risk quadrant. Its usage frequency is 180 times, exceeding the preset usage frequency threshold of 150 times. Based on the strategy, intermediate processing measures are initiated: confirming its low-risk status; prioritizing its inclusion in the core analysis of the current complex case, detailing how it was used to pinpoint key facts, and recording any minor corrections made after its use and their effects; prioritizing its use as key evidence in reporting and presentation sessions, with the system recording its specific role in case progression; simultaneously, implementing routine monitoring, comparing its monitoring data with that of high-value-high-risk evidence to establish risk trends; finally, the assessment deems additional protective measures necessary, restricting direct modification permissions at the system permission level, and recording and feeding back the entire processing flow and effects to the high-value-high-risk quadrant processing strategy library, providing a reference for corrective effects and monitoring models for future processing of similar high-value-low-risk but frequently used evidence.

[0099] like Figure 2The system architecture diagram of an intelligent management system applied to an evidence contamination platform is shown in the present invention. The intelligent management system for an evidence contamination platform includes:

[0100] Data Acquisition and Preprocessing Module: Includes a data acquisition unit and a data preprocessing unit; wherein, the data acquisition unit collects historical data from the platform; the data preprocessing unit preprocesses the collected historical data from the platform, identifies and processes missing values, detects and processes abnormal data points, deletes duplicate records, and standardizes and denoises the text data;

[0101] The evidence value assessment module includes an influence analysis unit, a statistical relationship analysis unit, and an evidence value score fusion unit. The influence analysis unit uses the PageRank algorithm to analyze evidence nodes and citation relationship networks, calculating the PageRank score for each evidence node. The statistical relationship analysis unit collects and organizes data on the number of times evidence is cited by top journals and its usage frequency within the platform, constructs a weighted linear regression model, analyzes the impact of these two factors on the evidence value, and generates an evidence value score. The evidence value score fusion unit weights and fuses the PageRank score and the weighted linear regression score according to preset weights to generate the final comprehensive evidence value score.

[0102] The pollution risk assessment module includes a structured risk extraction unit, a non-graph feature processing unit, and a risk score fusion unit. The structured risk extraction unit abstracts evidence entities, users, and their interactions into nodes and edges in a graph data structure, forming graph data. This graph data is then input into a graph neural network (GNN) model to capture and output a risk feature score based on the graph structure. The non-graph feature processing unit collects and processes the non-graph structural features of the evidence, trains a random forest classifier, determines whether the evidence poses a pollution risk, and outputs a risk score based on non-graph features. The risk score fusion unit merges the structured risk score with the non-graph feature risk score to generate the final comprehensive pollution risk score.

[0103] The evidence value-pollution risk coordinate system construction module includes a coordinate mapping unit and a quadrant division unit. The coordinate mapping unit receives the evidence value score and the pollution risk score, and maps each evidence point to a two-dimensional coordinate system, where the horizontal axis represents the evidence value score and the vertical axis represents the pollution risk score. The quadrant division unit divides the two-dimensional coordinate system into four quadrants according to a preset threshold: high value-high risk quadrant, high value-low risk quadrant, low value-high risk quadrant, and low value-low risk quadrant.

[0104] The evidence processing strategy decision-making module includes a priority and hierarchy determination unit, a threshold judgment and status confirmation unit, and a strategy matching and generation unit. The priority and hierarchy determination unit determines the processing priority order and level of evidence based on its quadrant in the coordinate system, according to preset priority rules and processing levels. The threshold judgment and status confirmation unit checks whether the specific attributes of evidence in a specific quadrant meet the prescribed threshold conditions, and for evidence that meets the conditions, confirms its specific status within that quadrant. The strategy matching and generation unit combines the evidence's quadrant, the determined processing level, and the results of the threshold judgment and status confirmation unit to accurately match and generate corresponding processing strategies from a predefined strategy library, generating specific, executable operations or suggestions, and implementing feedback and linkage between strategies.

[0105] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An intelligent management method applied to an evidence contamination platform, characterized in that, include: Historical data from the collection platform, including identified pollution events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals, are preprocessed. Based on preprocessed historical data, an evidence value assessment model and a pollution risk assessment model are constructed. The evidence value assessment model uses a weighted scoring fusion model algorithm, which analyzes influence ranking and citation relationships using the PageRank algorithm to identify high-impact evidence and extract node importance. It also uses a weighted linear regression algorithm to analyze the number of citations by top journals and the frequency of evidence use. The comprehensive evidence value score is calculated by combining the PageRank score and the linear regression weight score. The pollution risk assessment model constructs a graph data structure, uses the Graph Neural Network (GNN) algorithm to extract risk features, and uses the Random Forest algorithm to build a classification model for non-graph features. The comprehensive pollution risk score is generated by fusing the GNN output and the classification model results. The evidence value score and pollution risk score are mapped to a two-dimensional coordinate system, with the evidence value score as the horizontal axis and the pollution risk score as the vertical axis, to establish a dual-coordinate decision-making mechanism of evidence value and pollution risk. Different risk areas and thresholds are set, and the quadrants are divided into high-value-high-risk quadrant, high-value-low-risk quadrant, low-value-high-risk quadrant, and low-value-low-risk quadrant according to the update priority. Different processing strategies are assigned according to different quadrants.

2. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The historical data collected by the platform includes identified pollution events, frequency of evidence use, pollution source characteristics, pollution diffusion pathways, correction effects, influence rankings, and citation counts in top journals. The data undergoes preprocessing, including: Collect historical data from the evidence contamination platform, including identified contamination events, frequency of evidence use, pollution source characteristics, pollution diffusion paths, correction effects, influence rankings, and number of citations by top journals; The collected historical data from the platform is preprocessed. Based on the data type and distribution, missing values ​​are filled by the mean or mode. Outlier data points are identified and processed using the Z-score statistical method. Duplicate records are deleted. Regular expressions are used to standardize and denoise the text data.

3. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The evidence value assessment model uses a weighted scoring fusion model algorithm. It analyzes influence ranking and citation relationships using the PageRank algorithm to identify high-impact evidence and extract node importance. It then uses a weighted linear regression algorithm to analyze the number of citations by top journals and the frequency of evidence use, including: The PageRank algorithm is used to iteratively calculate the evidence nodes and their citation relationship network. Based on the quality and quantity of nodes entering the chain, the relative influence of each evidence node in the network is evaluated and quantified, generating a node importance score based on the network structure. The calculation formula is as follows: ； Where PR(A) is the PageRank score of evidence node A, representing the influence of evidence node A in the evidence network; d is the damping coefficient, representing the probability that the user will continue to click the link; 1-d is the complement of the damping coefficient, representing the probability that the user will not continue to click the link; B(A) is the set of all evidence nodes pointing to evidence node A; PR(T) is the PageRank score of evidence node T, representing the influence of evidence node T in the evidence network; C(T) is the number of backlinks of evidence node T. Collect and organize data on the number of times each piece of evidence is cited in top journals and its usage frequency within the platform, using the following formula: ； Where Y is the evidence value score, X1 is the number of times cited by top journals, X2 is the frequency of evidence use, β0 is the intercept, and β1 and β2 are regression coefficients, representing the degree of influence of each independent variable on the dependent variable. It is an error term; A weighted linear regression model is constructed, with evidential value as the dependent variable and the number of times the evidence is cited in top-tier journals and the frequency of use of the evidence as independent variables. Different weights are assigned based on the historical revision effect or influence ranking of the evidence. A regression equation is fitted, these two factors are quantified, and an evidential value score based on statistical relationships is generated. The formula is as follows: ； w i Y is the weight of the i-th sample, set according to the historical correction effect or influence ranking of the evidence, adjusting the importance of this sample in the regression model; i It is the actual value of the i-th sample; is the predicted value of the i-th sample; n is the number of samples; this formula finds the regression coefficients and fits the regression equation by minimizing the weighted sum of squared residuals; where the weighted linear regression score LR is equal to the predicted value of the regression model. .

4. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The calculation of the comprehensive evidentiary value score, which combines PageRank score and linear regression weight score, includes: The PageRank score PR(A) and the weighted linear regression score LR are weighted and fused according to preset weights to generate the final comprehensive evidence value score V, as shown in the formula: ； Where V is the final generated comprehensive evidence value score, representing the overall value of a certain evidence node; α is a weighting coefficient between 0 and 1, representing the proportion of the PageRank score PR(A) in the final score V; 1-α represents the proportion of the weighted linear regression score LR.

5. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The pollution risk assessment model constructs a graph data structure, uses the Graph Neural Network (GNN) algorithm to extract risk features, and uses the Random Forest algorithm to build a classification model for non-graph features, including: Evidence entities, users, and interaction relationships are abstracted into nodes and edges in a graph, forming graph data that can represent the flow and correlation of evidence. The constructed graph data is input into a graph neural network model. The GNN learns the representation of nodes in the graph structure layer by layer by aggregating the features of the nodes themselves and the features of the neighboring nodes, thereby capturing risk patterns. Collect and process the non-graph structure features of the evidence, and train a random forest classifier; the random forest determines whether the evidence has a risk of contamination by constructing multiple decision trees and combining the prediction results.

6. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The fusion of GNN output and classification model results generates a comprehensive pollution risk score, including: The structured risk features extracted by GNN are fused with the non-graph features utilized by random forest to comprehensively assess the contamination risk or probability of evidence. The formula is as follows: ； Wherein, FRS is the final integrated pollution risk score; GS is the pollution risk score output by the graph neural network model; RS is the pollution risk score output by the non-graph feature model; θ is a weighting coefficient used to adjust the relative importance of GS and RS, with a value ranging from 0 to 1.

7. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The process of mapping the evidence value score and pollution risk score to a two-dimensional coordinate system, with the evidence value score as the horizontal axis and the pollution risk score as the vertical axis, to establish a dual-coordinate decision-making mechanism based on evidence value and pollution risk includes: The evidence value score and pollution risk score are mapped onto a two-dimensional coordinate system. The horizontal axis X represents the evidence value score, reflecting the importance, relevance, and probative value of the evidence to the case. The higher the score, the greater the potential role of the evidence in the case. The vertical axis Y represents the pollution risk level score, reflecting the possibility or severity of the evidence being contaminated. The higher the score, the lower the reliability of the evidence and the greater the risk of misleading the judgment of the case. By plotting these two scores for each piece of evidence onto the coordinate system, an evidence point is formed. The specific location of the evidence point in the coordinate system reflects the overall value and risk of the evidence.

8. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The different risk zones and thresholds are set and categorized into high-value-high-risk quadrants, high-value-low-risk quadrants, low-value-high-risk quadrants, and low-value-low-risk quadrants according to update priority, including: Based on the established two-dimensional coordinate system of evidence value and pollution risk, thresholds are set for evidence value score and pollution risk score, and the four quadrants are set as high value-high risk quadrant, low value-high risk quadrant, low value-low risk quadrant and high value-low risk quadrant respectively. The high-value-high-risk quadrant is located in the upper right corner of the coordinate system. Evidence with both evidentiary value and contamination risk exceeds its respective preset threshold. This type of evidence has a significant impact on the case, but its reliability is questionable, and it requires the highest priority for attention and processing. The high-value-low-risk quadrant is located in the lower right corner of the coordinate system. Evidence with evidentiary value exceeds the preset evidentiary value threshold, while its contamination risk is below the preset contamination risk threshold. This type of evidence has a significant impact on the case but a low contamination risk, and requires the second priority for identification and processing. The low-value-high-risk quadrant is located in the upper left corner of the coordinate system. Evidence with evidentiary value is below the preset evidentiary value threshold, while its contamination risk is above the preset contamination risk threshold. This type of evidence has little interference but a high risk profile, and requires the third priority for identification and processing. The low-value-low-risk quadrant is located in the lower left corner of the coordinate system. Evidence with both evidentiary value and contamination risk are below their respective preset thresholds, and this type of evidence has the lowest priority for updating.

9. The intelligent management method for an evidence contamination platform according to claim 1, characterized in that, The allocation of different processing strategies according to different quadrants includes: For evidence located in the high-value-high-risk quadrant, when the frequency of use of the evidence exceeds a preset frequency threshold, the highest level of handling measures are initiated; comprehensive pollution source tracing and characteristic analysis are performed, and this information is compared with the platform's historical pollution event database to identify pollution sources and potential transmission patterns; cross-disciplinary experts are organized to assess the pollution impact, and the scope and depth of pollution are determined by referring to the correction effect records of similar pollution to low-value-high-risk evidence; mandatory pollution elimination or evidence isolation measures are implemented, and the entire process is recorded; For evidence located in the high-value-low-risk quadrant, when the frequency of use of such evidence exceeds a preset frequency threshold, intermediate processing measures are initiated; its low-risk status is confirmed, and it is prioritized for inclusion in case analysis, with its utilization and correction effects recorded; it is prioritized for presentation as key evidence, with its specific role and effect in case progression recorded; routine monitoring is implemented, and the monitoring data is compared and analyzed with the monitoring data of high-value-high-risk evidence to form a risk warning linkage; it is assessed whether additional protection measures are needed to prevent future contamination, and the processing procedures and effects are recorded and fed back to the high-value-high-risk quadrant for reference; For evidence located in the low-value-high-risk quadrant, when the frequency of use of the evidence exceeds a preset frequency threshold, basic processing measures are initiated; risk technology assessment is conducted to confirm the pollution status and high-risk attributes; pollution elimination is performed or the evidence is marked as unacceptable and excluded, with removal being given priority to reduce risk; the processing results are archived and fed back to the model update, while the processing strategy is used as a reference and linked to the decision-making process of the low-value-low-risk quadrant. For evidence located in the low-value-low-risk quadrant, when the frequency of use of the evidence is lower than a preset frequency threshold, minimum processing measures are initiated; after confirming low risk, a standardized archiving process is adopted, a low-value tag is added, and the evidence is retrieved under specific search requests to monitor its usage. The processing flow is then fed back to the high-value-low-risk quadrant.

10. An intelligent management system for an evidence contamination platform, using the intelligent management method for an evidence contamination platform as described in any one of claims 1-9, characterized in that, include: Data Acquisition and Preprocessing Module: Includes a data acquisition unit and a data preprocessing unit; wherein, the data acquisition unit collects historical data from the platform; the data preprocessing unit preprocesses the collected historical data from the platform, identifies and processes missing values, detects and processes abnormal data points, deletes duplicate records, and standardizes and denoises the text data; The evidence value assessment module includes an influence analysis unit, a statistical relationship analysis unit, and an evidence value score fusion unit. The influence analysis unit uses the PageRank algorithm to analyze evidence nodes and citation relationship networks, calculating the PageRank score for each evidence node. The statistical relationship analysis unit collects and organizes data on the number of times evidence is cited by top journals and its usage frequency within the platform, constructs a weighted linear regression model, analyzes the impact of these two factors on the evidence value, and generates an evidence value score. The evidence value score fusion unit weights and fuses the PageRank score and the weighted linear regression score according to preset weights to generate the final comprehensive evidence value score. The pollution risk assessment module includes a structured risk extraction unit, a non-graph feature processing unit, and a risk score fusion unit. The structured risk extraction unit abstracts evidence entities, users, and their interactions into nodes and edges in a graph data structure, forming graph data. This graph data is then input into a graph neural network (GNN) model to capture and output a risk feature score based on the graph structure. The non-graph feature processing unit collects and processes the non-graph structural features of the evidence, trains a random forest classifier, determines whether the evidence poses a pollution risk, and outputs a risk score based on non-graph features. The risk score fusion unit merges the structured risk score with the non-graph feature risk score to generate the final comprehensive pollution risk score. The evidence value-pollution risk coordinate system construction module includes a coordinate mapping unit and a quadrant division unit. The coordinate mapping unit receives the evidence value score and the pollution risk score, and maps each evidence point to a two-dimensional coordinate system, where the horizontal axis represents the evidence value score and the vertical axis represents the pollution risk score. The quadrant division unit divides the two-dimensional coordinate system into four quadrants according to a preset threshold: high value-high risk quadrant, high value-low risk quadrant, low value-high risk quadrant, and low value-low risk quadrant. The evidence processing strategy decision-making module includes a priority and hierarchy determination unit, a threshold judgment and status confirmation unit, and a strategy matching and generation unit. The priority and hierarchy determination unit determines the processing priority order and level of evidence based on its quadrant in the coordinate system, according to preset priority rules and processing levels. The threshold judgment and status confirmation unit checks whether the specific attributes of evidence in a specific quadrant meet the prescribed threshold conditions, and for evidence that meets the conditions, confirms its specific status within that quadrant. The strategy matching and generation unit combines the evidence's quadrant, the determined processing level, and the results of the threshold judgment and status confirmation unit to accurately match and generate corresponding processing strategies from a predefined strategy library, generating specific, executable operations or suggestions, and implementing feedback and linkage between strategies.

Images