Simulation scene optimization method and system based on interpretable machine learning algorithm

Abstract

The invention discloses a simulation scene optimization method and system based on an interpretable machine learning algorithm, and the method comprises two parts: factor importance analysis and equivalent test scene recognition, achieves the analysis and evaluation of the weight of a factor based on a simulation result, intelligently determines the factor composition used for generating a scene, recognizes an equivalent generated scene, and achieves the optimization of the simulation scene. And the purpose of simplifying the simulation scene set for evaluation is achieved. By simplifying the test scene, the evaluation efficiency of the intelligent model algorithm to be tested in the simulation combat scene is improved, the simulation effect is enhanced, the interpretability and confidence of the evaluation result are improved, and the iterative optimization speed of the intelligent algorithm is also improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent simulation, and in particular to a simulation scene optimization method and system based on an interpretable machine learning algorithm. Background Art

[0002] In the generation of existing intelligent simulation combat scenarios, massive and high-coverage test scenarios are usually generated based on numerous experimental factors. The intelligent model algorithm to be tested needs to complete simulation tests under all test scenarios to obtain reliable intelligent model algorithm evaluation results. However, the massive simulation test scenarios require a huge amount of computing resources for test evaluation, and also seriously affect the speed of algorithm iteration optimization. In order to solve this problem, it is necessary to streamline the simulation scenarios used for evaluation and find a smaller set of simulation scenarios with similar evaluation effects. Summary of the invention

[0003] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art and provide a simulation scenario optimization method and system based on an interpretable machine learning algorithm, which evaluates the importance of factors with the help of an interpretable machine learning algorithm and identifies equivalent test scenarios, thereby achieving optimization and simplification of the test scenarios while maintaining the evaluation effect.

[0004] The technical solution of the present invention is: a simulation scenario optimization method based on an interpretable machine learning algorithm, including two parts: factor importance analysis and equivalent test scenario identification;

[0005] The steps of factor importance analysis include:

[0006] S1. Generate a complete set of simulation test scenarios S according to given test factors and factor levels;

[0007] S2. Perform simulation deduction on the algorithm to be tested in all simulation test scenarios, obtain the simulation deduction results corresponding to each simulation test scenario, and form a simulation deduction result set R of the algorithm to be tested in the full set of simulation test scenarios S;

[0008] S3,<S,R> As the data set to be explained, based on the interpretable machine learning algorithm, the contribution of each experimental factor to the simulation results generated by the algorithm to be tested is calculated;

[0009] S4. According to the contribution of each test factor, the whole set S of simulation test scenarios is optimized to reduce the number of simulation test scenarios;

[0010] The step of identifying the equivalent test scenario comprises:

[0011] P1. Obtain the original value range of the given test factor; for each test factor, randomly generate multiple smaller value ranges within its original value range, and the combination of each smaller value range is the original value range; the smaller value ranges of different test factors are combined into different value ranges to generate different simulation test scenarios;

[0012] P2. In the simulation scenarios generated by different value range combinations, simulation tests are performed on the algorithms to be tested respectively, and the simulation results are recorded;

[0013] P3. Statistically analyze the simulation results to select the equivalent test factor value ranges in each smaller value range of each test factor;

[0014] P4. According to the value range of the equivalent test factor of each factor, the simulation test scenarios are optimized to reduce the number of simulation test scenarios.

[0015] Use both factor importance analysis and equivalent test scenario identification, or use either method, to optimize the simulation scenarios and reduce the number of simulation test scenarios.

[0016] Further, the S1 specifically includes:

[0017] Determine all experimental factors and factor levels to be analyzed;

[0018] Describe the factor levels of each experimental factor numerically;

[0019] One way of taking values ​​of all experimental factors constitutes a scenario, and each experimental factor takes values ​​in turn within its factor level range to form the complete set S of simulation test scenarios.

[0020] Furthermore, in S3, for the algorithm to be tested based on the tree model, TreeExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm, and for other types of algorithms to be tested, KernelExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm.

[0021] Further, the S3 specifically includes:

[0022] right<S,R> , using a random sampling method, generate multiple sub-datasets that contain part of the data set to be explained and are consistent with the distribution of the data set to be explained;

[0023] For each sub-dataset, the SHAP value of each experimental factor is calculated using the selected SHAP interpreter;

[0024] The SHAP values ​​of the same experimental factor in each sub-data set are averaged to obtain the global SHAP value of each experimental factor, that is, the contribution of each experimental factor to the simulation results generated by the tested algorithm.

[0025] Furthermore, in P2, a Monte Carlo simulation method is used to perform a simulation test.

[0026] The present invention also provides a simulation scenario optimization system based on an interpretable machine learning algorithm, comprising a factor importance analysis module and an equivalent test scenario identification module;

[0027] The factor importance analysis module includes:

[0028] The test scenario generation unit I is used to generate a complete set S of simulation test scenarios according to given test factors and factor levels;

[0029] A simulation deduction unit is used to simulate and deduce the algorithm to be tested in all simulation test scenarios, obtain the simulation deduction results corresponding to each simulation test scenario, and form a simulation deduction result set R of the algorithm to be tested in the full set S of simulation test scenarios;

[0030] Factor weight calculation unit, used to<S,R> As the data set to be explained, based on the interpretable machine learning algorithm, the contribution of each experimental factor to the simulation results generated by the algorithm to be tested is calculated;

[0031] The factor optimization unit is used to optimize the whole set S of simulation test scenarios according to the contribution of each test factor, so as to reduce the number of simulation test scenarios;

[0032] The equivalent test scene recognition module includes:

[0033] The test scenario generation unit II is used to obtain the original value range of the given test factor; for each test factor, within its original value range, a plurality of smaller value ranges are randomly generated, and the combination of each smaller value range is the original value range; the smaller value ranges of different test factors are combined into different value ranges to generate different simulation test scenarios;

[0034] A simulation test unit is used to perform simulation tests on the algorithms to be tested in simulation scenarios generated by different value range combinations and record the simulation results;

[0035] A scene recognition unit is used to perform statistical analysis on the simulation results and screen out the equivalent test factor value ranges in each smaller value range of each test factor;

[0036] The scenario optimization unit is used to optimize the simulation test scenario according to the value range of the equivalent test factor of each factor, so as to reduce the number of simulation test scenarios.

[0037] The factor importance analysis module and the equivalent test scenario identification module are used simultaneously, or either module is used to optimize the simulation scenario and reduce the number of simulation test scenarios.

[0038] Furthermore, in the test scenario generation unit I, the specific method of generating the complete set S of simulation test scenarios according to the given test factors and factor levels is: determining all the test factors and factor levels to be analyzed; digitally describing the factor level of each test factor; one value taking method of all the test factors constitutes a scenario, and each test factor takes values ​​in turn within its factor level range to constitute the complete set S of simulation test scenarios.

[0039] Furthermore, in the factor weight calculation unit, an optional SHAP interpreter is provided. For the algorithm to be tested based on a tree model, TreeExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm. For other types of algorithms to be tested, KernelExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm.

[0040] Furthermore, in the factor weight calculation unit, the specific method for determining the contribution size of each test factor is:<S,R> , using a random sampling method, generate multiple sub-datasets that contain part of the data set to be explained and have the same distribution as the data set to be explained; for each sub-dataset, use the selected SHAP interpreter to calculate the SHAP value of each experimental factor; average the SHAP values ​​of each experimental factor of each sub-dataset to obtain the global SHAP value of each experimental factor, that is, to obtain the contribution of each experimental factor to the simulation results generated by the algorithm to be tested.

[0041] Furthermore, in the simulation test unit, a Monte Carlo simulation method is used to perform the simulation test.

[0042] The advantages of the present invention compared with the prior art are:

[0043] (1) The present invention uses an interpretable machine learning algorithm to improve the interpretability and confidence of the evaluation results.

[0044] (2) The present invention reduces the number of test scenarios and improves the efficiency of test evaluation by reducing the number of test factors and optimizing the factor value range.

[0045] (3) The present invention uses fewer equivalent scenarios for test evaluation, thereby improving the speed of iterative optimization of the intelligent algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0047] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0048] The generation of intelligent simulation combat scenarios needs to consider many factors such as weather, terrain, combat forces, combat equipment, tactics and strategies, and each factor has a different range of values, resulting in a large number of intelligent simulation scenarios being generated, and the simulation training time being prolonged, which affects the practicality of intelligent simulation. In order to solve the above problems, the present invention provides a simulation scenario optimization method based on an interpretable machine learning algorithm, which includes two parts: factor importance analysis and equivalent test scenario identification. The specific process can be found in Figure 1 ,The two parts can be executed separately or successively according to the needs of scene streamlining.

[0049] 1. The steps of factor importance analysis include:

[0050] S1. Generate a complete set of simulation test scenarios S according to the given test factors and factor levels

[0051] A preferred implementation is:

[0052] (1) Obtaining experimental factors: First, it is necessary to determine all experimental factors and factor levels to be analyzed. These factors are used to generate various intelligent simulation combat scenarios.

[0053] (2) Data preprocessing: In the data preprocessing of factors, a factor can be regarded as a characteristic dimension, and then the factor level of each factor can be digitally described, and necessary normalization, standardization or other necessary preprocessing operations can be performed. A matrix evaluation method is adopted, and a value of all factors constitutes a scenario vector. Each factor takes different factor levels in turn to obtain the complete set S of simulation test scenarios for subsequent analysis and modeling.

[0054] The method for generating the complete set S of simulation test scenarios is not limited to the above method, and other methods may be selected to generate the complete set S of simulation test scenarios according to the actual test requirements of the algorithm to be tested.

[0055] S2. Simulate the algorithm to be tested in all simulation test scenarios to obtain the simulation results corresponding to each simulation test scenario, forming a simulation result set R of the algorithm to be tested in the full set of simulation test scenarios S.

[0056] S3,<S,R> As the data set to be explained, based on the interpretable machine learning algorithm, calculate the contribution of each experimental factor to the simulation results generated by the tested algorithm

[0057] A preferred implementation is:

[0058] (1) Select an interpretable model: Select an appropriate interpretable machine learning algorithm based on the type of intelligent model algorithm to be tested and the characteristics of the data. For tree-based models (such as decision trees, random forests, etc.), select TreeExplainer as the SHAP interpreter; for any other model (such as a complex neural network model), select KernelExplainer as the SHAP interpreter; in addition, for models that only require local explanations, select the local SHAP interpreter to perform interpretable analysis on the model.

[0059] (2) Sampling to generate sub-datasets: Use random sampling to generate multiple sub-datasets from the full set of simulation test scenarios S. Each sub-dataset should contain a portion of the data in the original data set, and the distribution of the sub-dataset should be consistent with that of the original data set.

[0060] (3) Calculate the SHAP value of the sub-dataset: For each sub-dataset, use the selected SHAP interpreter to calculate the SHAP values ​​of all factors. These values ​​represent the contribution of each factor to the model prediction results in the sub-dataset.

[0061] (4) Calculate the global SHAP value: The SHAP values ​​of all sub-datasets are averaged to obtain the global SHAP value of each factor. These values ​​represent the contribution of each factor to the model prediction results in the entire data set.

[0062] S4. According to the contribution of each test factor, the whole set of simulation test scenarios S is optimized to reduce the number of simulation test scenarios.

[0063] A preferred implementation is:

[0064] (1) Visualizing the global SHAP value: Visualize the calculated global SHAP value, such as drawing a factor importance bar chart, scatter plot, heat map, etc., to intuitively show the contribution of each factor to the prediction result.

[0065] (2) Analyze the global SHAP value: By analyzing the visualization results, we can understand which factors have the greatest impact on the prediction results in the entire data set and how they affect the prediction results.

[0066] (3) Optimize test factors: Based on the analysis results, optimize the test factors to improve the predictive performance and interpretability of the intelligent algorithm model. Specifically, unimportant test factors (contribution is less than the set value) can be removed based on the analysis results, thereby reducing the number of simulation test scenarios.

[0067] 2. The steps of equivalent test scenario identification include:

[0068] P1. Obtain the original value range or distribution pattern of the given test factor; for each test factor, randomly generate multiple smaller value ranges within its original value range, and the combination of each smaller value range is the original value range; the smaller value ranges of different test factors form different value range combinations, thereby generating different simulation test scenarios.

[0069] A preferred implementation is:

[0070] (1) Obtaining the initial value range: For each test factor generated in the test scenario, obtain its original value range. A probability model can be established to describe its value range and distribution law. This probability model should be able to reflect the uncertainty and variation range of the test factor.

[0071] (2) Randomly generate a smaller value range: For each test factor, a smaller value range is randomly generated within its original value range. Each smaller value range still constitutes a complete value range. The smaller value ranges of different test factors form different value range combinations, which in turn generate different simulation scenarios. The number of simulation scenarios here does not decrease. The generated value range combination should conform to the data distribution law of the test factor to which it belongs.

[0072] P2. In the simulation scenarios generated by different value range combinations, simulate the algorithms to be tested and record the simulation results.

[0073] A preferred implementation is:

[0074] Monte Carlo simulation test is used: the intelligent algorithm to be tested is tested and evaluated on the simulation scenarios generated by different value range combinations, the simulation results are recorded, and the changes and results of various parameters in the process are recorded.

[0075] Specifically, for the test and evaluation of the intelligent algorithm of the missile weapon system to be tested, this step specifically includes:

[0076] (1) Select the algorithm to be tested: First, select the object to be tested. For different algorithm types such as perception, navigation, decision-making, and attack, and algorithm functions such as target recognition, situation fusion, firepower allocation, and Qida attack, select the corresponding scenarios that can test the algorithm to be tested and the scenario change dimensions and levels to verify the intelligence of the algorithm. Load the intelligent algorithm into the simulation environment and configure the initial parameters according to the algorithm requirements to ensure that the algorithm can run normally.

[0077] (2) Select simulation scene elements: including scene background, geographical dimensions such as sea, land, air, and sky, weather, atmosphere, rain, snow, trees, and other environmental factors, and red and blue team equipment such as ships, missiles, and vehicles, to complete the selection of basic elements of the simulation scene for a specific algorithm type and function.

[0078] (3) Constructing a simulation test question bank: Based on the selected scenario elements, combined with the variable test factor selection and factor range value, a complete set of simulation test scenario question banks of different difficulty levels are constructed in a matrix manner.

[0079] (4) Construct an indicator system: Based on the type and function of the algorithm being tested, combined with the simulation test question bank, construct parameters that can qualitatively and quantitatively evaluate the algorithm during the test process, record the changes in key parameters during the algorithm execution process, including the algorithm's own performance parameters, functional parameters, equipment parameters affected by the algorithm's command output, such as missile trajectory, target tracking status, etc., as well as changes in the simulated battlefield process and result parameters. This is used as a parameter measurement for evaluating the equivalence of scenarios.

[0080] (5) Obtain simulation results: Through the visual interface, evaluate the impact of changes in different parameters under different scenarios on indicators and other parameters.

[0081] P3. Perform statistical analysis on the test evaluation results to screen out the equivalent test factor value ranges in each smaller value range of each test factor.

[0082] A preferred implementation is:

[0083] Statistical analysis of test results: Statistical analysis is performed on the test results of the test evaluation to calculate various statistics, such as mean, variance, standard deviation, etc. By analyzing these statistics, the equivalent range of test factor values ​​can be screened out.

[0084] Specifically, the test and evaluation of the missile weapon system intelligent algorithm includes:

[0085] (1) Simulation data collection and organization: Based on the constructed indicator system and the records of changes in key parameters during the algorithm execution process, including the values ​​of different test factors (such as algorithm parameters, environmental conditions, target characteristics, etc.) and the corresponding functional performance indicators (such as hit rate, response time, resource consumption, etc.), and corresponding to the selection of corresponding test factors and factor parameters in each simulation scenario, check and clean up errors, missing values ​​or outliers in the data to ensure the integrity and accuracy of the data.

[0086] (2) Statistical analysis: Calculate the mean, median, mode, variance, standard deviation and other statistics of each experimental factor and its corresponding performance indicator to provide basic information on data distribution, and analyze the correlation between the experimental factors and between the experimental factors and the performance indicators. The correlation coefficient (such as Pearson correlation coefficient and Spearman rank correlation coefficient) can be used for quantification.

[0087] (3) Screening equivalent value range:

[0088] a) Grouping and comparison: Based on the statistical results, the values ​​of the experimental factors are grouped. Cluster analysis (K-means clustering) or methods based on statistical differences (t-test, ANOVA analysis of variance) can be used to determine which values ​​belong to the same group or "equivalent" group statistically.

[0089] b) Setting thresholds: Based on the tolerance range of the simulation scenario test characteristics and algorithm function requirements, set thresholds for changes in one or more performance indicators. Any test factor value that changes within the threshold range can be considered equivalent.

[0090] c) Determine the equivalent range: Based on the grouping results and threshold settings, determine the equivalent value range of each test factor.

[0091] (4) Cross-validation and adjustment: The validity of the selected equivalent value range is verified by cross-validation using a test data set that does not participate in model training. The impact of small changes in the test factors on the simulation process parameters and simulation results within the equivalent value range is further analyzed.

[0092] P4. According to the value range of the equivalent test factor of each factor, optimize the simulation test scenario and reduce the number of simulation test scenarios.

[0093] A preferred implementation is:

[0094] Optimize the range of test factor values: Based on the analysis results, the range of equivalent test factor values ​​can be optimized. This process can include adjusting the upper and lower limits of the range, changing the distribution law of the range, etc., so that the range of equivalent test factor values ​​is closer to the actual test conditions, thereby improving the accuracy and reliability of the Monte Carlo method.

[0095] After the factor importance analysis and equivalent test scenario identification are completed, it is possible to verify whether the simplified scenario is equivalent to the original scenario, including: obtaining the optimized test factors and their optimized value ranges, that is, simplifying the matrix test factors from the row and column dimensions. Re-conduct the simulation test evaluation test of the intelligent model algorithm on the simplified scenario set, and verify whether the simplified scenario can achieve the equivalent test evaluation function as before the simplification based on the evaluation results and the changes in process parameters.

[0096] It is to be understood that the present invention is described by way of embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and embodiments that can fall within the scope of the claims of this application all fall within the scope protected by the present invention.

[0097] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A simulation scene optimization method based on an explainable machine learning algorithm, characterized in that: It includes two parts: factor importance analysis and equivalent test scenario identification; The steps of factor importance analysis include: S1. Generate a complete set of simulation test scenarios S according to given test factors and factor levels; S2. Perform simulation deduction on the algorithm to be tested in all simulation test scenarios, obtain the simulation deduction results corresponding to each simulation test scenario, and form a simulation deduction result set R of the algorithm to be tested in the full set of simulation test scenarios S; S3,<S,R> As the data set to be explained, based on the interpretable machine learning algorithm, the contribution of each experimental factor to the simulation results generated by the algorithm to be tested is calculated; S4. According to the contribution of each test factor, the whole set S of simulation test scenarios is optimized to reduce the number of simulation test scenarios; The step of identifying the equivalent test scenario comprises: P1. Obtain the original value range of the given test factor; for each test factor, randomly generate multiple smaller value ranges within its original value range, and the combination of each smaller value range is the original value range; the smaller value ranges of different test factors are combined into different value ranges to generate different simulation test scenarios; P2. In the simulation scenarios generated by different value range combinations, simulation tests are performed on the algorithms to be tested respectively, and the simulation results are recorded; P3. Statistically analyze the simulation results to select the equivalent test factor value ranges in each smaller value range of each test factor; P4. According to the value range of the equivalent test factor of each factor, the simulation test scenarios are optimized to reduce the number of simulation test scenarios; Use both factor importance analysis and equivalent test scenario identification, or use either method, to optimize the simulation scenarios and reduce the number of simulation test scenarios.

2. The simulation scene optimization method based on an explainable machine learning algorithm according to claim 1, characterized in that: The S1 specifically includes: Determine all experimental factors and factor levels to be analyzed; Describe the factor levels of each experimental factor numerically; One way of taking values ​​of all experimental factors constitutes a scenario, and each experimental factor takes values ​​in turn within its factor level range to form the complete set S of simulation test scenarios.

3. The simulation scene optimization method based on an explainable machine learning algorithm according to claim 1, characterized in that: In S3, for the algorithm to be tested based on the tree model, TreeExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm, and for other types of algorithms to be tested, KernelExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm.

4. The simulation scene optimization method based on an explainable machine learning algorithm according to claim 1, characterized in that: The S3 specifically includes: right<S,R> , using a random sampling method, generate multiple sub-datasets that contain part of the data set to be explained and are consistent with the distribution of the data set to be explained; For each sub-dataset, the SHAP value of each experimental factor is calculated using the selected SHAP interpreter; The SHAP values ​​of the same experimental factor in each sub-data set are averaged to obtain the global SHAP value of each experimental factor, that is, the contribution of each experimental factor to the simulation results generated by the tested algorithm.

5. The simulation scene optimization method based on an explainable machine learning algorithm according to claim 1, characterized in that: In P2, a Monte Carlo simulation method is used to perform simulation experiments.

6. A simulation scene optimization system based on an explainable machine learning algorithm, characterized in that: Includes factor importance analysis module and equivalent test scenario identification module; The factor importance analysis module includes: The test scenario generation unit I is used to generate a complete set S of simulation test scenarios according to given test factors and factor levels; A simulation deduction unit is used to simulate and deduce the algorithm to be tested in all simulation test scenarios, obtain the simulation deduction results corresponding to each simulation test scenario, and form a simulation deduction result set R of the algorithm to be tested in the full set S of simulation test scenarios; Factor weight calculation unit, used to<S,R> As the data set to be explained, based on the interpretable machine learning algorithm, the contribution of each experimental factor to the simulation results generated by the tested algorithm is calculated; The factor optimization unit is used to optimize the whole set S of simulation test scenarios according to the contribution of each test factor, so as to reduce the number of simulation test scenarios; The equivalent test scene recognition module includes: The test scenario generation unit II is used to obtain the original value range of the given test factor; for each test factor, within its original value range, a plurality of smaller value ranges are randomly generated, and the combination of each smaller value range is the original value range; the smaller value ranges of different test factors are combined into different value ranges to generate different simulation test scenarios; A simulation test unit is used to perform simulation tests on the algorithms to be tested in simulation scenarios generated by different value range combinations and record the simulation results; A scene recognition unit is used to perform statistical analysis on the simulation results and screen out the equivalent test factor value ranges in each smaller value range of each test factor; The scenario optimization unit is used to optimize the simulation test scenario according to the value range of the equivalent test factor of each factor, so as to reduce the number of simulation test scenarios. The factor importance analysis module and the equivalent test scenario identification module are used simultaneously, or either module is used to optimize the simulation scenario and reduce the number of simulation test scenarios.

7. The simulation scene optimization system based on an explainable machine learning algorithm according to claim 6, characterized in that: In the test scenario generation unit I, the specific method of generating the complete set S of simulation test scenarios according to the given test factors and factor levels is as follows: determining all the test factors and factor levels to be analyzed; digitally describing the factor level of each test factor; one way of taking values ​​of all the test factors constitutes a scenario, and each test factor takes values ​​in turn within its factor level range to constitute the complete set S of simulation test scenarios.

8. The simulation scene optimization system based on an explainable machine learning algorithm according to claim 6, characterized in that: In the factor weight calculation unit, an optional SHAP interpreter is provided. For the algorithm to be tested based on the tree model, TreeExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm. For other types of algorithms to be tested, KernelExplainer is used as the SHAP interpreter of the interpretable machine learning algorithm.

9. The simulation scene optimization system based on an explainable machine learning algorithm according to claim 6, characterized in that: In the factor weight calculation unit, the specific method for determining the contribution size of each experimental factor is:<S,R> , using a random sampling method, generate multiple sub-datasets that contain part of the data set to be explained and have the same distribution as the data set to be explained; for each sub-dataset, use the selected SHAP interpreter to calculate the SHAP value of each experimental factor; average the SHAP values ​​of each experimental factor of each sub-dataset to obtain the global SHAP value of each experimental factor, that is, to obtain the contribution of each experimental factor to the simulation results generated by the algorithm to be tested.

10. The simulation scene optimization method based on an explainable machine learning algorithm according to claim 1, characterized in that: In the simulation test unit, a Monte Carlo simulation method is used to perform simulation tests.

Images