基于IF-CM-LOF的尾矿坝位移监测数据离群值诊断

Outlier Diagnosis of Tailings Dam Displacement Monitoring Data Based on IF-CM-LOF

为解决孤立森林(IF)算法在离群值识别过程中对于边界位置数据处理结果的模糊性和不确定性问题,提高监测数据中异常值的检出率,在使用IF算法进行离群值初步识别的基础上,将IF量化计算后的异常得分引作变量,导入云模型(CM)逆向云发生器,根据逆向云变换所得的云数字特征值实现边界数据定位,进一步引入局部异常因子(LOF)算法对所定位的边界数据进行二次精确诊断,以某尾矿坝地表位移监测数据为例进行了模型验证。研究结果表明:对于监测数据中真实异常值和边界部分随机误差,IF模型检出率为16.5%和22.2%,而IF-CM-LOF模型的检出率分别达到90%和61.1%,离群值诊断性能明显优于IF模型。

In order to solve the problems of fuzzle and uncertainty in the processing results of boundary position data by isolated forest(IF)algorithm in the process of outlier identification,and improve the detection rate of outliers in monitoring data,on the basis of using the IF algorithm for preliminary identification of outliers,the outlier scores after quantitative calculation were introduced into the cloud model(CM)reverse cloud generator as variables.Based on the cloud digital eigenvalues obtained by the reverse cloud transform,the boundary data was located.The local anomaly factor(LOF)algorithm was further introduced to make the secondary accurate diagnosis of the located boundary data.The surface displacement monitoring data of a tailings dam was taken as an example to verify the model.The results show that for the real outliers and boundary random errors in the monitoring data,the detection rates of the IF model are 16.5%and 22.2%,while the detection rates of the IF-CM-LOF model are 90%and 61.1%,respectively.The diagnostic performance of outliers is obviously better than that of the IF model.