Machine learning-based comprehensive evaluation method and system for environmental remediation performance and safety of micro-nano iron-based reducing materials

Abstract

This invention discloses a machine learning-based method and system for comprehensively evaluating the environmental remediation performance and safety of micro / nano iron-based reducing materials. The method includes: acquiring environmental remediation performance and safety scores predicted using machine learning; introducing a bottleneck penalty mechanism—when each score is below a corresponding basic compliance threshold, a linear penalty function is used to significantly amplify the weight of low-scoring indicators; introducing a high-zone attenuation mechanism—when each score is above a corresponding high-zone boundary, a saturation function is used to reduce the marginal contribution rate of ultra-high-scoring indicators; calculating the dynamic weight corresponding to each score based on the outputs of the bottleneck penalty and high-zone attenuation mechanisms; and obtaining a comprehensive evaluation score by weighted summation of the environmental remediation performance and safety scores using the dynamic weights. This significantly improves R&D efficiency and safety, and its dynamic fusion and evaluation methodology is highly universal, extending to other functional material evaluation fields requiring the balance of multiple indicators, such as catalysis and pharmaceuticals.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary technical field of environmental materials assessment and artificial intelligence decision support, specifically involving a comprehensive assessment method and system for the environmental remediation performance and safety of micro-nano iron-based reducing materials based on machine learning. Background Technology

[0002] Micro- and nano-iron-based reducing agents are widely used in soil and groundwater remediation due to their high specific surface area, efficient electron transport capabilities, and strong reducing activity. However, with the deepening of their application, their potential environmental risks are gradually emerging. In complex environmental media transformation processes, micro- and nano-iron-based materials may induce the generation of reactive oxygen species or exert toxic effects on non-target organisms by releasing metal ions; in addition, the materials may interact with target pollutants and their degradation intermediates, triggering combined toxicity risks. Therefore, in environmental remediation engineering, a comprehensive performance-safety assessment of micro- and nano-iron-based reducing agents has become a common industry requirement.

[0003] Despite the increasing importance of comprehensive assessment, existing technologies still have significant limitations in their assessment methodologies, primarily in the following two aspects. First, assessment models based on traditional experimental simulations are inefficient and difficult to quantify. Existing assessment methods largely rely on physical simulations under laboratory conditions. This trial-and-error approach is not only time-consuming, costly, and resource-intensive, but also struggles to encompass the vast combinations of material modification parameters and pollutant types in real-world environments. More importantly, traditional experimental methods cannot accurately quantify the independent contribution weights of various influencing factors (such as material structural characteristics and environmental factors) to remediation effects or toxicity responses under multivariate conditions, resulting in limited guidance for rational material design. Second, performance evaluation and safety evaluation are disconnected, lacking a unified decision-making basis. Currently, the technical field typically treats them as independent, parallel indicators: one focuses on the degradation kinetics and mechanisms of specific pollutants under simulated conditions in performance research; the other emphasizes the toxicological responses of materials to typical model organisms in the safety dimension. This discrete assessment model fails to resolve the inherent contradictions between the two dimensions. For example, a certain material may exhibit extremely high pollutant removal rates, but at the same time, it may be accompanied by a high risk of biotoxicity (high performance - high risk); conversely, a highly safe material may have low remediation efficiency (low performance - low risk). Existing technologies have failed to provide a scientific, systematic, and quantitative method to weigh these two key dimensions, causing decision-makers to often rely on subjective experience when faced with multiple technology options, which severely restricts the scientific nature of material selection and the efficiency of optimization design.

[0004] Currently, while machine learning has been used to predict single indicators, greatly alleviating the difficulties of experimental simulation methods in handling high-dimensional, heterogeneous data and mining nonlinear relationships, there is still no existing technology that can nonlinearly and dynamically fuse dual-domain measurement models. In other words, there is a lack of a comprehensive evaluation algorithm or indicator system that can reflect the synergistic effect and inherent trade-off between performance and safety. Therefore, there is an urgent need in this field for a breakthrough comprehensive evaluation method that can scientifically integrate multi-dimensional prediction results, quantify the balance between remediation efficiency and environmental risk, and thus provide integrated intelligent decision support for the research, screening, and practical application of micro / nano iron-based materials. Summary of the Invention

[0005] In view of the above, to address the core problem of the separation between performance evaluation and safety assessment in existing technologies and the lack of a unified quantitative comprehensive evaluation standard, the purpose of this invention is to provide a machine learning-based method and system for comprehensively evaluating the environmental remediation performance and safety of micro / nano iron-based reducing materials. This method dynamically weights and fuses the two independent predictive scores of remediation performance evaluation and safety assessment to generate a unified, comprehensive, and clearly instructive comprehensive evaluation score.

[0006] To achieve the above-mentioned objectives, this embodiment provides a comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning. This method innovatively combines two independent prediction scores—performance and safety—with dynamic weighting and fusion to generate a unified and comparable comprehensive evaluation score. Specifically, it includes the following steps: Obtain environmental remediation performance and safety scores predicted using machine learning methods; A weak link penalty mechanism is introduced: when the score of each category (safety / performance) is lower than the corresponding basic compliance threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function. A high-zone attenuation mechanism is introduced: when the score of each category is higher than the corresponding high-zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function. The dynamic weights corresponding to each type of score are calculated based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; the environmental remediation performance score and the safety score are weighted and summed using the dynamic weights to obtain the comprehensive evaluation score.

[0007] Preferably, a weakness penalty mechanism is introduced: when the score for each category is lower than the corresponding basic threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function, including: The linear penalty function is ,in, Score the i-th type of indicator. This represents the basic threshold for achieving the target score for the i-th type of indicator. This is a performance penalty factor used to construct the dynamic weights for fusion; when < Calculate the performance penalty factor. , ≥ ,but = 1, indicating no penalty.

[0008] Preferably, a high-zone attenuation mechanism is introduced. When the score of each category is higher than the corresponding high-zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function, including: The saturation function is ,in Score the i-th type of indicator. The high partition boundary corresponding to the score of the i-th type of indicator. This is the kurtosis coefficient of the function. This is a performance degradation factor used to construct the dynamic weights for fusion; when > At that time, the Sigmoid function ensures that the score enters... After the above high partitioning, its positive contribution to dynamic weights increases more gradually. ≤ ,but = 1 indicates no decay, reflecting the principle of diminishing marginal utility.

[0009] Preferably, the dynamic weights corresponding to each type of score are calculated based on the output of the weakest link penalty mechanism and the high-area attenuation mechanism, including: First, calculate the original weighted pressure: in, This represents the performance penalty factor corresponding to the i-th type of indicator score output by the bottleneck penalty mechanism. This represents the performance degradation factor corresponding to the i-th type of index score output by the high-area degradation mechanism. This represents the original weight pressure of the i-th category indicator score; Then, the normalization calculation is performed to determine the dynamic weight corresponding to each type of score; Where i and j are both rating indices. These represent the scores corresponding to environmental remediation performance and safety, respectively. This represents the dynamic weight of the score for the i-th type of indicator.

[0010] Preferably, a comprehensive evaluation score is obtained by weighting and summing the environmental remediation performance score and the safety score using dynamic weights, including: in, This represents the overall evaluation score. and These represent the environmental remediation performance score and its corresponding dynamic weight, respectively. and These represent the security score and its corresponding dynamic weight, respectively.

[0011] Preferably, for the environmental remediation performance score, the corresponding basic compliance threshold and high-level boundary are 0.8 and 0.9, respectively; for the safety score, the corresponding basic compliance threshold and high-level boundary are 0.8 and 0.9, respectively.

[0012] Preferably, the environmental remediation performance score is a probability value between 0 and 1 output by a performance binary classification (degradable / non-degradable) prediction model based on the input characteristics of micro / nano iron-based reducing materials, target pollutant characteristics, and environmental condition characteristics; the safety score is a probability value between 0 and 1 output by a safety binary classification (increased toxicity / decreased toxicity) prediction model based on the input characteristics of micro / nano iron-based reducing materials, target pollutant characteristics, environmental condition characteristics, and test biological characteristics.

[0013] To achieve the above-mentioned objectives, this invention also provides a machine learning-based comprehensive evaluation system for the environmental remediation performance and safety of micro / nano iron-based reducing materials. This system dynamically weights and fuses two independent prediction scores—performance and safety—to generate a unified and comparable comprehensive evaluation score, including: The data acquisition module is used to acquire environmental remediation performance scores and safety scores predicted using machine learning methods. The weakness penalty mechanism module is used to introduce a weakness penalty mechanism. When the score of each category is lower than the corresponding basic threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function. The high-zone attenuation mechanism module is used to introduce a high-zone attenuation mechanism. When the score of each category is higher than the corresponding high zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function. The dynamic weighting module is used to calculate the dynamic weights corresponding to each type of score based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; the environmental remediation performance score and the safety score are weighted and summed by the dynamic weights to obtain the comprehensive evaluation score.

[0014] To achieve the above-mentioned objectives, the embodiments also provide a computing device, including a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the above-mentioned comprehensive evaluation method for the environmental remediation performance and safety of micro-nano iron-based reducing materials based on machine learning.

[0015] To achieve the above-mentioned objectives, the embodiments also provide a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements the above-mentioned comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning.

[0016] Compared with the prior art, the beneficial effects of the present invention include at least the following: This invention pioneers a comprehensive evaluation framework that integrates degradation performance and composite toxicity, enabling rapid prediction of any material-pollutant-environment combination based on machine learning. By introducing a unique dynamic weighting function that incorporates a bottleneck penalty and high-zone attenuation mechanism, it scientifically quantifies the trade-off between performance and safety, solving the problem of discrete indicators in existing technologies and achieving a leap from multi-indicator presentation to single-indicator decision-making. This framework significantly improves R&D efficiency and safety, and its dynamic fusion and dynamic evaluation methodology is highly universal, extendable to other functional material evaluation fields requiring the balance of multiple indicators, such as catalysis and pharmaceuticals. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning, provided in the embodiment. Figure 2 These are contour plots of the dynamically weighted results under different performance / security scores provided in the embodiments; Figure 3 This is a schematic diagram of the structure of the comprehensive evaluation system for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning provided in the embodiment. Figure 4 This is an output image of the comprehensive evaluation system for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning, provided in the embodiment. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0020] The inventive concept of this invention is to address the problem of separate evaluation of remediation performance and environmental safety in existing technologies, and the lack of unified quantitative standards. This invention provides a machine learning-based comprehensive evaluation scheme for the environmental remediation performance and safety of micro / nano iron-based reducing materials. First, a machine learning model is used to predict the probability scores of pollutant degradation performance and biosafety under specific environmental conditions. Then, a dynamic weighted comprehensive evaluation function is constructed and applied. By setting a basic compliance threshold and a high-zone boundary, a short-board penalty mechanism is introduced to significantly amplify the weight of low-scoring indicators, and a high-zone attenuation mechanism is introduced to reduce the marginal contribution rate of ultra-high-scoring indicators. Finally, the dynamic weights are calculated, and a single comprehensive evaluation score is generated.

[0021] like Figure 1 As shown in the embodiment, a comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning is provided, including the following steps: S1, obtain the environmental remediation performance score and safety score predicted using machine learning methods.

[0022] In this embodiment, the environmental remediation performance score is based on the output of a performance prediction model, which predicts the output environmental remediation performance score based on the combined characteristics of the input micro / nano iron-based reducing materials, pollutants, and the environment. The safety score is based on the output of a safety prediction model, which predicts the output safety score based on the combined characteristics of the input micro / nano iron-based reducing materials, pollutants, the environment, and the test organism. The specific construction process includes: 1. Dataset Construction: An environmental remediation performance dataset for micro / nano iron-based reducing materials was established. It covers material properties (e.g., purity, average particle size, surface modification type), pollutant properties (e.g., initial pollutant concentration, molecular structure), and reaction environmental conditions (pH, temperature). Labels were set as binary variables: degradable and non-degradable. A biocomplex toxicity dataset was also established, which, in addition to the above features, includes the classification of the tested organisms, photoperiod, and exposure time. Labels were set as binary variables: increased toxicity and decreased toxicity.

[0023] 2. Feature Engineering and Preprocessing: PaDEL-Descriptor software was used to calculate two-dimensional (2D) molecular descriptors for materials and pollutants to quantify their microscopic chemical structure characteristics. The raw data was cleaned, including missing value imputation, one-heat encoding conversion of categorical features, and standardization of numerical features. Principal component analysis was then used to reduce the dimensionality of the high-dimensional molecular descriptors, forming standardized input feature vectors.

[0024] Specifically, discrete features (such as modification type and biological species) are converted into one-hot encoding; features with a missing value rate exceeding 30% are removed, and the remaining missing values ​​are imputed using the median method. For high-dimensional molecular descriptor features, principal component analysis (PCA) is used for dimensionality reduction, retaining principal components with a cumulative variance contribution rate exceeding 85%; all numerical features are Z-score standardized to eliminate dimensional differences.

[0025] 2. Model Training and Ensemble. Using the AutoGluon automated machine learning framework, performance prediction and security prediction models were trained separately based on the dataset. The training process employed a multi-layer stacked ensemble strategy, integrating multiple models such as gradient boosting trees, random forests, and K-nearest neighbors.

[0026] 3. Model Validation and Application: After completing model training and confirming through cross-validation and independent test sets that the model possesses the preset high accuracy and generalization ability, the core evaluation indicators were used, including the area under the receiver operating characteristic curve (ROC-AUC), accuracy, and F1 score. Validation results show that the model is robust and generalizable. The scientific validity of the model was also verified through interpretability analysis. Based on the validated model, it was then deployed. The performance prediction model receives the combined characteristics of the material to be evaluated, the target pollutant, and environmental conditions as input, and calculates and outputs a probability score of the pollutant degradation performance (i.e., environmental remediation performance). The biosafety prediction model takes as input the combined characteristics of the material to be evaluated, contaminants, environmental conditions, and the test organism, and calculates and outputs a probability score of biosafety. The probability scores are all continuous values ​​between 0 and 1, thus providing a standardized quantitative data foundation for subsequent dynamic weighted comprehensive evaluation.

[0027] S2 introduces a weakness penalty mechanism: when the score of each category is lower than the corresponding basic threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function.

[0028] In this embodiment, a basic compliance threshold and a high-level boundary are set for each type of indicator score. Specifically, for the environmental remediation performance score, the corresponding basic compliance threshold and high-level boundary are 0.8 and 0.9, respectively; for the safety score, the corresponding basic compliance threshold and high-level boundary are 0.8 and 0.9, respectively. These basic compliance thresholds and high-level boundaries are derived from domain expert knowledge and are used to define the ranges of qualified (meeting the basic compliance threshold), good (between the basic compliance threshold and the high-level boundary), and excellent (above the high-level boundary).

[0029] When any score falls below its corresponding baseline threshold, a bottleneck penalty mechanism is introduced. This mechanism uses a linear penalty function to significantly amplify the weight of low-scoring indicators, thereby significantly increasing the relative importance of that score in subsequent weight calculations. This forces the evaluation system to focus on and amplify the system's bottleneck effect. Specifically, this includes: The linear penalty function is ,in, Score the i-th type of indicator. This represents the basic threshold for achieving the target score for the i-th type of indicator. This is used as a performance penalty factor to increase the proportion of this weakest performance indicator in the final weighting, and the proportion increases as the deviation from the threshold increases. Used to construct dynamic weights for fusion; when < Calculate the performance penalty factor. , ≥ ,but = 1, indicating no penalty. This weakest link penalty mechanism ensures that any failure to meet the standard will have a significant impact on the final weight.

[0030] S3 introduces a high-zone attenuation mechanism: when the score of each category is higher than the corresponding high-zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function.

[0031] When any score exceeds the preset high-scoring threshold, a high-scoring attenuation mechanism is introduced. This mechanism uses a saturation function (such as the sigmoid function) to reduce the marginal contribution rate of ultra-high-scoring indicators, preventing the extreme high scores of a single indicator from masking the shortcomings of another indicator and encouraging a balanced development of performance and security. Specifically, this includes: The saturation function is ,in Score the i-th type of indicator. The high partition boundary corresponding to the score of the i-th type of indicator. (=0.7) is the maximum attenuation coefficient. (=25) is the kurtosis coefficient of the function. This is a performance degradation factor used to construct the dynamic weights for fusion; when > At that time, the Sigmoid function ensures that the score enters... After the above high partitioning, its positive contribution to dynamic weights increases more gradually. ≤ ,but = 1 indicates no decay, reflecting the principle of diminishing marginal utility.

[0032] S4 calculates the dynamic weights corresponding to each type of score based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; and obtains the comprehensive evaluation score by weighted summation of the environmental remediation performance score and the safety score through the dynamic weights.

[0033] In this embodiment, the original weighted pressure is first calculated: in, This represents the performance penalty factor corresponding to the i-th type of indicator score output by the bottleneck penalty mechanism. This represents the performance degradation factor corresponding to the i-th type of index score output by the high-area degradation mechanism. This represents the original weight pressure of the i-th category indicator score; Then, the normalization calculation is performed to determine the dynamic weight corresponding to each type of score; Where i and j are both rating indices. These represent the scores corresponding to environmental remediation performance and safety, respectively. This represents the dynamic weight of the score for the i-th type of indicator; Finally, a comprehensive evaluation score is obtained by weighting and summing the environmental remediation performance score and the safety score using dynamic weights. in, This represents the overall evaluation score. and These represent the environmental remediation performance score and its corresponding dynamic weight, respectively. and These represent the security score and its corresponding dynamic weight, respectively.

[0034] Based on the above dynamic weighting method, a comprehensive evaluation of performance and security results is conducted, and the scores are as follows: Figure 2 As shown, the contour lines tending to be parallel to the coordinate axes at both ends reflect the effects of the high-area attenuation factor and the short-board penalty, while the contour lines bulging towards the middle indicate the friendliness of this dynamic weighted scoring to materials with balanced performance. For easier understanding, an example is provided below: Suppose there are two candidate materials, A and B: Material A: = 0.96 (Excellent) = 0.70 (Unacceptable); Material B: = 0.85 (Good) = 0.85 (Good); Under traditional linear weighting, A scores (0.96 + 0.70) / 2 = 0.83, and B scores (0.85 + 0.85) / 2 = 0.85, with B slightly better. However, under the dynamic weighting method of this invention, for material A, If the score is significantly below the threshold of 0.80, a weak point penalty will be triggered. The high-zone attenuation is triggered, and the final weighted score is (0.96×0.136+0.70×0.864)×100=73.5. For material B, both indicators are above the threshold and it does not enter the high zone. The penalty and attenuation factors are both about 1, and the weights are close to evenly distributed, so its score will be 85.

[0035] The results clearly demonstrate that the method of the present invention can effectively identify and penalize candidate solutions with obvious shortcomings (Material A), and tend to select solutions with balanced development (Material B), which is highly consistent with the risk avoidance principle in practical applications.

[0036] The above method, through the nonlinear and dynamic fusion of dual-model prediction results, achieves a leap from "multi-index presentation" to single-index decision-making. It can scientifically quantify the trade-off between performance and safety, effectively identify material shortcomings and encourage balanced development, and provide efficient and scientific decision support for high-throughput screening and optimized design of environmental remediation materials.

[0037] like Figure 3 As shown in the embodiment, an embodiment also provides a machine learning-based comprehensive evaluation system 40 for the environmental remediation performance and safety of micro / nano iron-based reducing materials, including a data acquisition module 41, a short-board penalty mechanism module 42, a high-zone attenuation mechanism module 43, and a dynamic weighting module 44. The data acquisition module 41 acquires the environmental remediation performance score and safety score predicted using machine learning methods; the short-board penalty mechanism module 42 introduces a short-board penalty mechanism that significantly amplifies the weight of low-scoring indicators through a linear penalty function when each score is below the corresponding basic compliance threshold; the high-zone attenuation mechanism module 43 introduces a high-zone attenuation mechanism that reduces the marginal contribution rate of ultra-high-scoring indicators through a saturation function when each score is above the corresponding high-zone boundary; the dynamic weighting module 44 calculates the dynamic weight corresponding to each score based on the outputs of the short-board penalty mechanism and the high-zone attenuation mechanism; and obtains a comprehensive evaluation score by weighted summation of the environmental remediation performance score and safety score using the dynamic weights.

[0038] It should be noted that the comprehensive evaluation device for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning provided in the above embodiments should be illustrated using the above-described functional module division as an example when conducting comprehensive safety evaluations. The functions can be assigned to different functional modules as needed, i.e., the internal structure of the terminal or server can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the comprehensive evaluation device for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning provided in the above embodiments and the embodiment of the comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning belong to the same concept. For details of its implementation process, please refer to the embodiment of the comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning, which will not be repeated here.

[0039] Based on the same inventive concept, the embodiment also provides a computing device, including a memory and one or more processors. The memory stores executable code, and when the one or more processors execute the executable code, it is used to implement the above-mentioned comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning, specifically including the following steps: S1, obtain the environmental remediation performance score and safety score predicted using machine learning methods; S2 introduces a weakness penalty mechanism: when the score of each category is lower than the corresponding basic threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function. S3 introduces a high-zone attenuation mechanism: when the score of each category is higher than the corresponding high-zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function. S4 calculates the dynamic weights corresponding to each type of score based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; and obtains the comprehensive evaluation score by weighted summation of the environmental remediation performance score and the safety score through the dynamic weights.

[0040] The computing device provided in this embodiment, at the hardware level, includes not only a processor and memory, but also internal buses, network interfaces, memory, and other hardware required for business operations. The memory is non-volatile memory. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning, as described in S1-S4 above. Of course, besides software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0041] Based on the same inventive concept, the embodiments also provide a computer-readable storage medium storing a program that, when executed by a processor, implements the aforementioned machine learning-based comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials. Figure 4 This is the result after the program is executed. Specifically, it includes the following steps: S1, obtain the environmental remediation performance score and safety score predicted using machine learning methods; S2 introduces a weakness penalty mechanism: when the score of each category is lower than the corresponding basic threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function. S3 introduces a high-zone attenuation mechanism: when the score of each category is higher than the corresponding high-zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function. S4 calculates the dynamic weights corresponding to each type of score based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; and obtains the comprehensive evaluation score by weighted summation of the environmental remediation performance score and the safety score through the dynamic weights.

[0042] In this embodiment, the computer-readable medium includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data.

[0043] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning, characterized in that, The two independent prediction scores, performance and safety, are dynamically weighted and fused to generate a unified and comparable comprehensive evaluation score. This process includes the following steps: Obtain environmental remediation performance and safety scores predicted using machine learning methods; A weak link penalty mechanism is introduced: when the score of each category (safety / performance) is lower than the corresponding basic compliance threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function. A high-zone attenuation mechanism is introduced: when the score of each category is higher than the corresponding high-zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function. The dynamic weights corresponding to each type of score are calculated based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; the environmental remediation performance score and the safety score are weighted and summed using the dynamic weights to obtain the comprehensive evaluation score.

2. The comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning as described in claim 1, characterized in that, A weakness penalty mechanism is introduced: when the score for each category falls below the corresponding basic threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function, including: The linear penalty function is ,in, Score the i-th type of indicator. This represents the basic threshold for achieving the target score for the i-th type of indicator. This is a performance penalty factor used to construct the dynamic weights for fusion; when < Calculate the performance penalty factor. , ≥ ,but = 1, indicating no penalty.

3. The comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning as described in claim 1, characterized in that, A high-zone attenuation mechanism is introduced: when the score of each category exceeds the corresponding high-zone boundary, the marginal contribution rate of the ultra-high-score index is reduced through a saturation function, including: The saturation function is ,in Score the i-th type of indicator. The high partition boundary corresponding to the score of the i-th type of indicator. This is the kurtosis coefficient of the function. The maximum attenuation coefficient, This is a performance degradation factor used to construct the dynamic weights for fusion; when > At that time, the Sigmoid function ensures that the score enters... After the above high partitioning, its positive contribution to dynamic weights increases more gradually. ≤ ,but = 1 indicates no decay, reflecting the principle of diminishing marginal utility.

4. The comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning according to claim 1, characterized in that, The dynamic weights corresponding to each type of score are calculated based on the output of the weakest link penalty mechanism and the high-zone attenuation mechanism, including: First, calculate the original weighted pressure: in, This represents the performance penalty factor corresponding to the i-th type of indicator score output by the bottleneck penalty mechanism. This represents the performance degradation factor corresponding to the i-th type of index score output by the high-area degradation mechanism. This represents the original weight pressure of the i-th category indicator score; Then, the normalization calculation is performed to determine the dynamic weight corresponding to each type of score; Where i and j are both rating indices. These represent the scores corresponding to environmental remediation performance and safety, respectively. This represents the dynamic weight of the score for the i-th type of indicator.

5. The comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning according to claim 1, characterized in that, A comprehensive evaluation score is obtained by weighting and summing the environmental remediation performance score and the safety score using dynamic weights, including: in, This represents the overall evaluation score. and These represent the environmental remediation performance score and its corresponding dynamic weight, respectively. and These represent the security score and its corresponding dynamic weight, respectively.

6. The comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning according to claim 1, characterized in that, For the environmental remediation performance score, the corresponding basic compliance threshold and high-level boundary are 0.8 and 0.9, respectively; for the safety score, the corresponding basic compliance threshold and high-level boundary are 0.8 and 0.9, respectively.

7. The comprehensive evaluation method for the environmental remediation performance and safety of micro / nano iron-based reducing materials based on machine learning according to claim 1, characterized in that, The environmental remediation performance score is a probability value between 0 and 1 output by the performance binary classification (degradable / non-degradable) prediction model based on the input characteristics of micro / nano iron-based reducing materials, target pollutant characteristics, and environmental condition characteristics; the safety score is a probability value between 0 and 1 output by the safety binary classification (increased toxicity / decreased toxicity) prediction model based on the input characteristics of micro / nano iron-based reducing materials, target pollutant characteristics, environmental condition characteristics, and test biological characteristics.

8. A machine learning-based comprehensive evaluation system for the environmental remediation performance and safety of micro / nano iron-based reducing materials, characterized in that, The two independent prediction scores, performance and safety, are dynamically weighted and fused to generate a unified and comparable comprehensive evaluation score, which includes: The data acquisition module is used to acquire environmental remediation performance scores and safety scores predicted using machine learning methods. The weakest link penalty mechanism module is used to introduce a weakest link penalty mechanism. When the score of each category (safety / performance) is lower than the corresponding basic compliance threshold, the weight of the low-scoring indicator is significantly amplified through a linear penalty function. The high-zone attenuation mechanism module is used to introduce a high-zone attenuation mechanism. When the score of each category is higher than the corresponding high zone boundary, the marginal contribution rate of the ultra-high score index is reduced through a saturation function. The dynamic weighting module is used to calculate the dynamic weights corresponding to each type of score based on the output of the short-board penalty mechanism and the high-zone attenuation mechanism; the environmental remediation performance score and the safety score are weighted and summed by the dynamic weights to obtain the comprehensive evaluation score.

9. A computing device comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the one or more processors execute the executable code, they are used to implement the comprehensive evaluation method for environmental remediation performance and safety of micro-nano iron-based reducing materials based on machine learning, as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a program that, when executed by a processor, implements the comprehensive evaluation method for the environmental remediation performance and safety of micro-nano iron-based reducing materials based on machine learning, as described in any one of claims 1-7.

Images