《Table 2AUCs with respect to different combinations of subsampling size and the number of isolation

  提示：宽带有限、当前游客访问压缩模式

本系列图表出处文件名：随高清版一同展现

《Application of isolation forest to extract multivariate anomalies from geochemical exploration data》

1. 获取 高清版本忘记账户？点击这里登录

1. 下载图表忘记账户？点击这里登录

[1]The number of isolation trees is 100；[2]subsampling size is 256.

Subsampling size and the number of isolation trees are the two essential parameters for the isolation forest algorithm in outlier detection.In order to test how the subsampling size and the number of isolation trees affect the performance of the isolation forest algorithm in geochemical anomaly identification，we first denoted subsampling size as value of 256 and the number of isolation trees as values of 1，5，10，20，50，100，150，200，250，and 300；and then we denoted the number of isolation trees as value of 100 and subsampling size as values of 16，32，64，128，256，512，1 028，2 048，4 096，and 8 192.We thus obtained twenty groups of parameters，each of which was used to initialize the isolation forest algorithm.The initialized model was applied to identify multivariate anomalies from the stream sediment data of the study area.Anomaly scores with respect to each group of parameters were computed and transformed into 200 by187 grid data using the Golden Software Surfer.The AUC values with respect to each group of parameters were estimated and listed in Table 2.The diagrams of AUC varying with subsampling size and the number of isolation trees were plotted and shown in Fig.4.