A method for identifying outliers in massive engineering detection data based on model integration
In order to improve the efficiency and accuracy of identifying outliers in a large amount of experimental detection data in highway engineering,an anomaly processing method"manual intervention+inspection model"was proposed.Through the comparison of four types of algorithms,including domain-based+probability-based+lin-ear-based+ensemble-based,the initial model was screened,and the LSCP ensemble idea was used to perform en-semble operation on the primary model,so as to realize the secondary classification of the detection dataset,improve the accuracy and robustness of model detection,and test the effectiveness of the model ensemble processing meth-od.The results showed that the model can achieve accuracy and sensitivity of outlier identification to 99.65% and 96.99%,respectively,and the detection performance is much higher than that of other single algorithms.The model algorithm was applied to the monitoring data detection of water inrush accident in a highway tunnel in Yunnan,and the average accuracy was 98.55%,indicating that the algorithm has a good anomaly detection effect.
highway engineeringdetection datamodel integrationidentification of abnormal pointsaccuracy
万方数据
维普
