论文部分内容阅读
准确的绝缘节破损预测能够保证铁路运输安全和经济效益。支持向量机算法能够处理轨道电路测试数据,对其进行分类,预测可能存在隐患的绝缘节,但支持向量机预测模型的原始样本多有冗余,基于此,提出了一种基于粗糙集和支持向量机的绝缘节破损预测模型。通过改进主分量启发式属性约简算法,降低样本维数,同时选用模拟退火算法完成SVM自动参数选优。实例分析和仿真结果表明,与单一支持向量机算法相比,属性约简后的粗糙集-支持向量机算法提高了分类器的分类性能,与采用网格搜索技术的SVM预测方法相比,模拟退火算法有效提高了SVM的预测精度。
Accurate insulation section damage prediction can ensure the safety and economic benefits of railway transportation. Support Vector Machine (SVM) algorithm can process orbital circuit test data to classify it and predict the potential insulation faults. However, the original samples of SVM prediction model are redundant. Based on this, a new algorithm based on rough set and support Prediction Model of Insulation Festival Damage of Vector Machine. By improving the principal component heuristic attribute reduction algorithm, the number of samples is reduced, and Simulated Annealing Algorithm is used to select SVM automatic parameters. The results of the example analysis and simulation show that compared with single support vector machine (SVM) algorithm, the attribute-reduced Rough Set-Support Vector Machine (SVM) algorithm improves the classification performance of the classifier. Compared with the SVM method based on grid search, Annealing algorithm effectively improves the prediction accuracy of SVM.