docs/00-manuscripts/ieee/reviews/review2/cover-letter-review2.md

In pat-s/2019-feature-selection: Research Compendium package for the publication "Monitoring forest health using hyperspectral imagery: Does feature selection improve the performance of machine-learning techniques?"

Dear Editor,

Please find the replies to all reviewer suggestions/questions in their respective files (labeled as reviewer<x>.md where <x> is the reviewer number starting from 1). Also we have added a changelog section further down below summarizing all changes we have applied to the manuscript as a result of this review.

With respect to the many helpful remarks of both reviewers, we did not implement/agree on the following suggested changes:

• The use of smoothing techniques as a preprocessing task for the hyperspectral data.
• Adding the spectral response function of the hyperspectral sensor.
• The inapplicability of using a buffer to reduce the influence of non-vegetation pixels and accounting for possible geometric offsets.
• A potentially inconsistent numbering of Figures / Tables.
• The request to include wrapper methods for feature selection.

We have outlined our reasons in detail in the respective in-line replies for each individual point.

In summary, we believe and hope to have addressed most reviewer remarks and that we incorporated a lot of helpful ideas from the reviewers into the current version of the manuscript and hope the manuscript can now be considered for publication.

Changelog

• Updated captions of almost all table and figure captions according to feedback from reviewer2
• Clarified the presence of inconsistent (= non 5% interval) defoliation values for plot Laukiz2
• Updated the study area map according to feedback from reviewer2
• Added a featureless learner to the benchmark results table as a baseline reference for performance comparison
• Updated the legend ordering and color palette in Fig. 3 according to feedback from reviewer2
• Rephrased the section discussion the performance of models which used PCA as a filter after feedback from reviewer2
• Clarified the removal of features with a pairwise correlation of 1 after feedback from reviewer2
• Added a reference why RBF Kernel was chosen as the kernel for the SVM learner after feedback from reviewer1
• Revised the use of abbreviations after feedback from reviewer2
• Rephrased the usefulness of machine learning for this study
• Added references to paragraph P.2, 2nd column on LL.47-5 as suggested by reviewer 1
• Added references analyzing defoliation caused by Diplodia Sapinea to section "Comparison to other studies"
• Pruned sections "Model differences" and "Performance vs. plot characteristics" to reduce potential overlaps with the "Results" section as indicated by reviewer2

pat-s/2019-feature-selection documentation built on Dec. 24, 2021, 8:40 a.m.