Folder: CHDC File: IDS.py 1. Chinese：’BEPCD模型的可解释性分析’ English：Interpretability analysis of the BEPCD model 2. Chinese：’使用SHAP进行特征重要性分析和模型决策解释’ English：Feature importance analysis and model decision explanation using SHAP 3. Chinese：’创建保存结果的目录’ English：Create directories to save results 4. Chinese：’已创建输出目录’ English：Output directories created 5. Chinese：’载入数据’ English：Load data 6. Chinese：’载入数据集并进行预处理’ English：Load dataset and preprocess 7. Chinese：’正在载入数据...’ English：Loading data... 8. Chinese：’载入或训练模型’ English：Load or train models 9. Chinese：’载入已保存的模型或训练新模型’ English：Load saved models or train new models 10. Chinese：’正在载入已训练的模型...’ English：Loading trained models... 11. Chinese：’正在训练模型...’ English：Training models... 12. Chinese：’SHAP特征重要性分析’ English：SHAP feature importance analysis 13. Chinese：’使用SHAP分析模型的特征重要性’ English：Analyze model feature importance using SHAP 14. Chinese：’正在分析{model_name}模型的特征重要性...’ English：Analyzing feature importance of {model_name} model... 15. Chinese：’模型的整体特征重要性’ English：Overall feature importance of the model 16. Chinese：’模型的特征影响详图’ English：Detailed feature impact plot of the model 17. Chinese：’模型对{class_name}类别的特征重要性’ English：Feature importance of the model for {class_name} class 18. Chinese：’与SHAP值的依赖关系’ English：Dependence with SHAP value 19. Chinese：’比较不同模型的特征重要性’ English：Compare feature importance of different models 20. Chinese：’比较Blending和Stacking模型的特征重要性...’ English：Comparing feature importance of Blending and Stacking models... 21. Chinese：’特征’ English：Feature 22. Chinese：’Blending重要性’ English：Blending importance 23. Chinese：’Stacking重要性’ English：Stacking importance 24. Chinese：’差异’ English：Difference 25. Chinese：’Blending模型的前10个重要特征’ English：Top 10 important features of the Blending model 26. Chinese：’Stacking模型的前10个重要特征’ English：Top 10 important features of the Stacking model 27. Chinese：’SHAP值重要性’ English：SHAP value importance 28. Chinese：’对{class_name}类别的特征重要性比较’ English：Feature importance comparison for {class_name} class 29. Chinese：’样本数量’ English：Number of samples 30. Chinese：’分析BEPCD模型的决策过程’ English：Analyze the decision process of the BEPCD model 31. Chinese：’分析BEPCD决策过程...’ English：Analyzing BEPCD decision process... 32. Chinese：’模型一致’ English：Models agree 33. Chinese：’Blending模型更有信心 ({b_prob:.2f} > {s_prob:.2f})’ English：Blending model is more confident ({b_prob:.2f} > {s_prob:.2f}) 34. Chinese：’Stacking模型更有信心 ({s_prob:.2f} > {b_prob:.2f})’ English：Stacking model is more confident ({s_prob:.2f} > {b_prob:.2f}) 35. Chinese：’实际标签’ English：Actual label 36. Chinese：’Blending预测’ English：Blending prediction 37. Chinese：’Stacking预测’ English：Stacking prediction 38. Chinese：’BEPCD预测’ English：BEPCD prediction 39. Chinese：’决策原因’ English：Reason for decision 40. Chinese：’BEPCD决策原因分布’ English：Distribution of BEPCD decision reasons 41. Chinese：’预测正确’ English：Prediction correct 42. Chinese：’不同决策原因的准确率’ English：Accuracy for different decision reasons 43. Chinese：’准确率’ English：Accuracy 44. Chinese：’类别的决策原因分布’ English：Distribution of decision reasons for the class 45. Chinese：’样本数’ English：Number of samples 46. Chinese：’正确预测’ English：Correct predictions 47. Chinese：’决策原因分布’ English：Distribution of decision reasons 48. Chinese：’类别的决策原因分布’ English：Distribution of decision reasons for {name} class 49. Chinese：’生成特征重要性和决策分析的综合报告’ English：Generate a comprehensive report of feature importance and decision analysis 50. Chinese：’生成分析报告...’ English：Generating analysis report... 51. Chinese：’BEPCD模型可解释性分析报告’ English：BEPCD Model Interpretability Analysis Report 52. Chinese：’特征重要性分析’ English：Feature Importance Analysis 53. Chinese：’整体特征重要性’ English：Overall Feature Importance 54. Chinese：’下表显示了Blending和Stacking模型的前10个最重要特征：’ English：The table below shows the top 10 most important features of the Blending and Stacking models: 55. Chinese：’模型特征重要性差异’ English：Differences in model feature importance 56. Chinese：’以下特征在Blending和Stacking模型之间表现出最大差异：’ English：The following features show the greatest differences between the Blending and Stacking models: 57. Chinese：’各攻击类型特征依赖’ English：Feature dependencies for each attack type 58. Chinese：’正常流量主要由CAN ID和数据字段的一致性模式特征区分。’ English：Normal traffic is mainly distinguished by the consistency pattern features of CAN ID and data fields. 59. Chinese：’DOS攻击主要依赖于CAN ID和数据字段值的异常检测，特别是短时间内大量相同CAN ID的出现。’ English：DOS attacks mainly rely on anomaly detection of CAN ID and data field values, especially the appearance of a large number of identical CAN IDs in a short period. 60. Chinese：’Fuzzing攻击对DataOne-DataFour的依赖性最高，这些字段的随机性是检测的关键。’ English：Fuzzing attacks rely most on DataOne-DataFour, and the randomness of these fields is key to detection. 61. Chinese：’重放攻击严重依赖时间特征和连续消息之间的关系，Before和Back特征对检测尤为重要。’ English：Replay attacks heavily rely on time features and the relationship between consecutive messages; Before and Back features are especially important for detection. 62. Chinese：’欺骗攻击主要通过CAN ID的组合和异常数据值检测，表现为合法但内容异常的消息。’ English：Spoofing attacks are mainly detected by combinations of CAN IDs and abnormal data values, appearing as legitimate but content-abnormal messages. 63. Chinese：’BEPCD决策过程分析’ English：BEPCD Decision Process Analysis 64. Chinese：’决策原因分布’ English：Distribution of decision reasons 65. Chinese：’BEPCD模型在做决策时的依据分布：’ English：The basis distribution when the BEPCD model makes decisions: 66. Chinese：’各类别的性能和决策特点’ English：Performance and decision characteristics of each class 67. Chinese：’主要决策依据:’ English：Main decision basis: 68. Chinese：’结论和建议’ English：Conclusions and Recommendations 69. Chinese：’主要发现’ English：Main Findings 70. Chinese：’改进建议’ English：Improvement Suggestions 71. Chinese：’CAN ID和DLC是区分正常和异常流量的最关键特征’ English：CAN ID and DLC are the most critical features for distinguishing normal and abnormal traffic 72. Chinese：’DataOne和DataTwo对检测特定类型的攻击（特别是Fuzzing和Spoofing）至关重要’ English：DataOne and DataTwo are crucial for detecting specific types of attacks (especially Fuzzing and Spoofing) 73. Chinese：’时间相关特征(Before和Back)在检测重放攻击方面表现出较高的重要性，但当前模型对这类攻击检测率仍有待提高’ English：Time-related features (Before and Back) show high importance in detecting replay attacks, but the current model's detection rate for such attacks still needs improvement 74. Chinese：’Blending模型更多依赖于CAN ID和单个数据字段，而Stacking模型对时间相关特征和连续消息之间的关系更敏感’ English：The Blending model relies more on CAN ID and individual data fields, while the Stacking model is more sensitive to time-related features and relationships between consecutive messages 75. Chinese：’这些差异解释了为什么BEPCD通过结合这些模型可以提高性能’ English：These differences explain why BEPCD can improve performance by combining these models 76. Chinese：’对重放攻击的检测可以通过增强时间特征工程得到改善，考虑添加更多时间序列特征’ English：Detection of replay attacks can be improved by enhancing time feature engineering and considering adding more time series features 77. Chinese：’考虑增加特征，如消息频率分析和顺序模式分析，以提高对各类攻击的检测能力’ English：Consider adding features such as message frequency analysis and sequential pattern analysis to improve detection capability for various attacks 78. Chinese：’针对低重放攻击检测率，可以考虑在BEPCD中加入专门针对时间序列异常的模型’ English：For low replay attack detection rates, consider adding models specifically for time series anomalies in BEPCD 79. Chinese：’优化模型选择机制，根据输入特征的特点智能选择最适合的模型进行预测’ English：Optimize the model selection mechanism to intelligently choose the most suitable model for prediction based on the characteristics of input features 80. Chinese：’增加对边缘案例的分析，特别是那些Blending和Stacking模型预测不一致的情况’ English：Increase the analysis of edge cases, especially those where Blending and Stacking model predictions are inconsistent 81. Chinese：’分析完成！所有结果已保存到./shap_results和./figures目录’ English：Analysis complete! All results have been saved to ./shap_results and ./figures directories Folder: CHDC File: BEPCD_analysis.py 1. Chinese：’BEPCD模型性能分析脚本’ English：BEPCD model performance analysis script 2. Chinese：’用于评估BEPCD模型的MAC值（模型平均复杂度）、单条数据识别时间和参数量’ English：Used to evaluate the MAC value (Model Average Complexity), single data recognition time, and parameter count of the BEPCD model 3. Chinese：’定义流式处理对象，用于逐一读取数据（模拟实时流分析）’ English：Define a streaming processing object for reading data one by one (simulate real-time stream analysis) 4. Chinese：’迭代处理pandas DataFrame/Series数据’ English：Iterate over pandas DataFrame/Series data 5. Chinese：’参数:’ English：Parameters: 6. Chinese：’特征数据’ English：Feature data 7. Chinese：’标签数据 (可选)’ English：Label data (optional) 8. Chinese：’返回:’ English：Returns: 9. Chinese：’逐行迭代X和y的生成器’ English：Generator for iterating X and y row by row 10. Chinese：’模拟BEPCD函数，用于性能分析’ English：Simulated BEPCD function for performance analysis 11. Chinese：’BEPCD模型预测函数’ English：BEPCD model prediction function 12. Chinese：’测试特征数据’ English：Test feature data 13. Chinese：’测试标签数据’ English：Test label data 14. Chinese：’Blending模型’ English：Blending model 15. Chinese：’Stacking模型’ English：Stacking model 16. Chinese：’每个类别的最佳模型列表’ English：Best model list for each class 17. Chinese：’真实标签列表’ English：True label list 18. Chinese：’预测标签列表’ English：Predicted label list 19. Chinese：’如果没有提供model参数，则构建一个默认的’ English：If the model parameter is not provided, build a default one 20. Chinese：’假定所有类别都是用blending_clf’ English：Assume all classes use blending_clf 21. Chinese：’对每个测试样本进行预测’ English：Predict for each test sample 22. Chinese：’BEPCD决策逻辑’ English：BEPCD decision logic 23. Chinese：’检查每个预测模型预测的类别是否由对应的模型作为最佳模型’ English：Check whether the predicted class of each model is the best model for that class 24. Chinese：’使用具有更高置信度的模型的预测’ English：Use the prediction of the model with higher confidence 25. Chinese：’加载保存的模型’ English：Load saved models 26. Chinese：’模型保存的目录路径’ English：Directory path where models are saved 27. Chinese：’尝试从文件加载模型’ English：Try to load models from file 28. Chinese：’成功加载模型文件’ English：Model files loaded successfully 29. Chinese：’如果模型文件不存在，则返回None’ English：If model files do not exist, return None 30. Chinese：’模型文件不存在或无法加载’ English：Model files do not exist or cannot be loaded 31. Chinese：’加载测试数据集’ English：Load test dataset 32. Chinese：’数据集路径’ English：Dataset path 33. Chinese：’测试特征数据’ English：Test feature data 34. Chinese：’测试标签数据’ English：Test label data 35. Chinese：’尝试从CSV文件加载数据’ English：Try to load data from CSV file 36. Chinese：’分离特征和标签’ English：Separate features and labels 37. Chinese：’除最后一列外的所有列作为特征’ English：All columns except the last as features 38. Chinese：’最后一列作为标签’ English：The last column as label 39. Chinese：’模拟测试集分割’ English：Simulate test set split 40. Chinese：’成功加载数据: {len(X_test)}行, {X_test.shape[1]}列’ English：Data loaded successfully: {len(X_test)} rows, {X_test.shape[1]} columns 41. Chinese：’加载数据失败: {str(e)}’ English：Failed to load data: {str(e)} 42. Chinese：’计算模型参数量’ English：Calculate model parameter count 43. Chinese：’机器学习模型对象’ English：Machine learning model object 44. Chinese：’参数总数’ English：Total parameter count 45. Chinese：’对于投票分类器，计算每个基模型的参数’ English：For voting classifiers, calculate parameters for each base model 46. Chinese：’随机森林’ English：Random forest 47. Chinese：’每个节点需要特征索引、阈值和子节点指针’ English：Each node requires feature index, threshold, and child node pointers 48. Chinese：’这是一个估计值’ English：This is an estimated value 49. Chinese：’默认返回特征数乘以一个系数’ English：By default, return feature count times a coefficient 50. Chinese：’计算模型的MAC值（模型平均复杂度）’ English：Calculate the MAC value (Model Average Complexity) of the model 51. Chinese：’输入特征数据’ English：Input feature data 52. Chinese：’MAC值’ English：MAC value 53. Chinese：’对于VotingClassifier’ English：For VotingClassifier 54. Chinese：’为每个基模型计算操作’ English：Calculate operations for each base model 55. Chinese：’每棵树需要评估所有特征’ English：Each tree needs to evaluate all features 56. Chinese：’其他模型的默认估计’ English：Default estimate for other models 57. Chinese：’分析BEPCD模型性能’ English：Analyze BEPCD model performance 58. Chinese：’性能分析结果字典’ English：Performance analysis result dictionary 59. Chinese：’用于分析的样本数量’ English：Number of samples for analysis 60. Chinese：’性能分析结果字典’ English：Performance analysis result dictionary 61. Chinese：’对X_test进行采样’ English：Sample X_test 62. Chinese：’1. 计算模型MAC值...’ English：1. Calculating model MAC value... 63. Chinese：’额外的决策复杂度’ English：Additional decision complexity 64. Chinese：’2. 计算模型参数量...’ English：2. Calculating model parameter count... 65. Chinese：’3. 计算单条数据处理时间...’ English：3. Calculating single data processing time... 66. Chinese：’测量BEPCD单条数据处理时间’ English：Measure BEPCD single data processing time 67. Chinese：’BEPCD预测过程’ English：BEPCD prediction process 68. Chinese：’简化版决策逻辑’ English：Simplified decision logic 69. Chinese：’转换为毫秒’ English：Convert to milliseconds 70. Chinese：’计算处理时间统计’ English：Calculate processing time statistics 71. Chinese：’4. 计算模型准确性...’ English：4. Calculating model accuracy... 72. Chinese：’性能分析结果字典’ English：Performance analysis result dictionary 73. Chinese：’可视化分析结果’ English：Visualize analysis results 74. Chinese：’参数: ‘ English：Parameters: 75. Chinese：’性能分析结果字典’ English：Performance analysis result dictionary 76. Chinese：’MAC值比较’ English：MAC value comparison 77. Chinese：’参数量比较’ English：Parameter count comparison 78. Chinese：’处理时间分布’ English：Processing time distribution 79. Chinese：’准确性指标’ English：Accuracy metrics 80. Chinese：’打印结果摘要’ English：Print result summary 81. Chinese：’打印结果摘要’ English：Print result summary 82. Chinese：’参数: ‘ English：Parameters: 83. Chinese：’性能分析结果字典’ English：Performance analysis result dictionary 84. Chinese：’BEPCD模型性能分析摘要’ English：BEPCD Model Performance Analysis Summary 85. Chinese：’1. MAC值分析 (模型平均复杂度)’ English：1. MAC Value Analysis (Model Average Complexity) 86. Chinese：’Blending模型 MAC值: {results['mac_value']['blending']:,} 操作’ English：Blending model MAC value: {results['mac_value']['blending']:,} operations 87. Chinese：’Stacking模型 MAC值: {results['mac_value']['stacking']:,} 操作’ English：Stacking model MAC value: {results['mac_value']['stacking']:,} operations 88. Chinese：’BEPCD模型 总MAC值: {results['mac_value']['bepcd']:,} 操作’ English：BEPCD model total MAC value: {results['mac_value']['bepcd']:,} operations 89. Chinese：’2. 模型参数量分析’ English：2. Model Parameter Count Analysis 90. Chinese：’Blending模型参数量: {results['parameters_count']['blending']:,} 参数’ English：Blending model parameter count: {results['parameters_count']['blending']:,} parameters 91. Chinese：’Stacking模型参数量: {results['parameters_count']['stacking']:,} 参数’ English：Stacking model parameter count: {results['parameters_count']['stacking']:,} parameters 92. Chinese：’BEPCD模型总参数量: {results['parameters_count']['bepcd']:,} 参数’ English：BEPCD model total parameter count: {results['parameters_count']['bepcd']:,} parameters 93. Chinese：’3. 单条数据处理时间分析’ English：3. Single Data Processing Time Analysis 94. Chinese：’平均处理时间: {results['processing_time']['average_ms']:.3f} 毫秒/样本’ English：Average processing time: {results['processing_time']['average_ms']:.3f} ms/sample 95. Chinese：’处理时间标准差: {results['processing_time']['std_ms']:.3f} 毫秒’ English：Processing time standard deviation: {results['processing_time']['std_ms']:.3f} ms 96. Chinese：’最小处理时间: {results['processing_time']['min_ms']:.3f} 毫秒’ English：Minimum processing time: {results['processing_time']['min_ms']:.3f} ms 97. Chinese：’最大处理时间: {results['processing_time']['max_ms']:.3f} 毫秒’ English：Maximum processing time: {results['processing_time']['max_ms']:.3f} ms 98. Chinese：’95%分位处理时间: {results['processing_time']['p95_ms']:.3f} 毫秒’ English：95th percentile processing time: {results['processing_time']['p95_ms']:.3f} ms 99. Chinese：’4. 模型准确性指标’ English：4. Model Accuracy Metrics 100. Chinese：’准确率: {results['accuracy']['accuracy']:.4f}’ English：Accuracy: {results['accuracy']['accuracy']:.4f} 101. Chinese：’精确率: {results['accuracy']['precision']:.4f}’ English：Precision: {results['accuracy']['precision']:.4f} 102. Chinese：’召回率: {results['accuracy']['recall']:.4f}’ English：Recall: {results['accuracy']['recall']:.4f} 103. Chinese：’F1分数: {results['accuracy']['f1']:.4f}’ English：F1 score: {results['accuracy']['f1']:.4f} 104. Chinese：’5. 性能总结’ English：5. Performance Summary 105. Chinese：’BEPCD理论处理速度: {samples_per_second:.2f} 样本/秒’ English：BEPCD theoretical processing speed: {samples_per_second:.2f} samples/second 106. Chinese：’处理1000条数据估计需要: {1000/samples_per_second:.2f} 秒’ English：Estimated time to process 1000 samples: {1000/samples_per_second:.2f} seconds 107. Chinese：’计算效率: {mac_per_time/1e9:.4f} GOPS (10亿操作/秒)’ English：Computation efficiency: {mac_per_time/1e9:.4f} GOPS (billion operations/second) 108. Chinese：’主函数’ English：Main function 109. Chinese：’BEPCD模型性能分析’ English：BEPCD Model Performance Analysis 110. Chinese：’1. 加载模型...’ English：1. Loading models... 111. Chinese：’如果模型加载失败，尝试创建模型’ English：If model loading fails, try to create models 112. Chinese：’模型加载失败，尝试创建模型...’ English：Model loading failed, trying to create models... 113. Chinese：’此处需要添加模型创建代码’ English：Model creation code needs to be added here 114. Chinese：’此脚本需要预先训练好的模型。请先运行IDS.ipynb训练模型并保存。’ English：This script requires pre-trained models. Please run IDS.ipynb to train and save the models first. 115. Chinese：’2. 加载测试数据...’ English：2. Loading test data... 116. Chinese：’数据加载失败，退出程序。’ English：Data loading failed, exiting program. 117. Chinese：’3. 分析BEPCD模型性能...’ English：3. Analyzing BEPCD model performance... 118. Chinese：’4. 打印结果摘要’ English：4. Printing result summary 119. Chinese：’5. 生成可视化结果...’ English：5. Generating visualization results... 120. Chinese：’6. 保存分析结果...’ English：6. Saving analysis results... 121. Chinese：’将numpy数组转换为列表以便JSON序列化’ English：Convert numpy arrays to lists for JSON serialization 122. Chinese：’分析完成！结果已保存到 'bepcd_performance_results.json' 和 'bepcd_performance_analysis.png'‘ English：Analysis complete! Results have been saved to 'bepcd_performance_results.json' and 'bepcd_performance_analysis.png' Folder: CHDC File: IDS. .ipynb Chinese：类别 English: CLASS Folder: Other model Test File: Ex.ipynb 1. Chinese：’# 使用训练好的模型对测试数据集X_test进行预测，得到预测结果y_pred。’ English：Use the trained model to predict the test dataset X_test and get the prediction result y_pred. 2. Chinese：’# 打印分类报告，其中包括每个类别的精确率、召回率、F1分数和支持度。’ English：Print the classification report, including precision, recall, F1 score, and support for each class. 3. Chinese：’# 对测试集进行预测并输出模型性能指标’ English：Predict on the test set and output model performance metrics. 4. Chinese：’# 计算并保存F1分数’ English：Calculate and save the F1 score. 5. Chinese：’# 计算真实标签y_test和预测标签y_pred之间的混淆矩阵。’ English：Calculate the confusion matrix between the true labels y_test and predicted labels y_pred. 6. Chinese：’# 绘制混淆矩阵’ English：Plot the confusion matrix.