针对单一机器学习模型在预测设备故障率的应用场景中存在误差大、精度低的问题,提出一种基于改进Stacking融合模型对铁路信号设备进行故障率预测的方法.采用XGBoost、LightGBM、CatBoost和逻辑回归方法构建基本Stacking模型,在此基础上引入Prophet时间序列预测模型,将Prophet模型提取的时序特征与基本Stacking模型逐级融合,构建改进后的Stacking-Prophet预测模型.最后以某单位信号设备数据为例,验证预测模型有效性.实验结果表明,相较单一预测模型,Stacking-Prophet预测模型均方根误差(RMSE)平均降低了 94.14%,预测精度有较大的提升,对设备运维有一定的参考价值.
To address the problems oflarge errors and low accuracy with single machine learning mod-els for predicting the failure rate of equipment,a prediction method based on improved Stacking fusion model is proposed.The basic Stacking fusion model is constructed by selecting XGBoost,LightGBM,Cat-Boost and the logistic regression model.On this basis,the Prophet time series prediction model is intro-duced,and the features extracted by the Prophet model are fused with the basic Stacking model level by level to construct the improved Stacking-Prophet prediction model.Finally,the validity of the prediction model is verified by taking the signal equipment data of a unit as an example.The experimental result shows that compared with the single prediction model,the Stacking-Prophet prediction model reduces the root mean square error(RMSE)by 94.14% on average,and the prediction accuracy is greatly improved.It is of a certain reference value for equipment operation and maintenance.