Anomaly Detection of Metallurgical Energy Data Based on iForest-AE

Round 1

Reviewer 1 Report

The authors have done a thorough study of the effectiveness of forest-AE model to detect anomalies in data. The experimental methods used have been explained in detail for the readers and results have been shown using a valid dataset. I recommend publishing this article after the following questions have been addressed by the authors:

1. Elaborate why SVM or the Forest, Autoencoder models alone cannot achieve the F1 Score or performance of a forest+autoencoder model - explain how the Forest-AE model removes outliers from the training data better than the other models? This will be useful to enable future development of improved models based on this study.

2. The IForest model alone has a low precision while the AE model has alone has high precision and moderate accuracy (Table 4). The authors should elaborate on why the AE or IForest models separately have worse performance than the combined IForest-AE model? How and why is the new algorithm able to achieve better performance than either the AE or IForest models or even the SVM model?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Refer to attachment

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

It is accepted for publication.