docs/00-manuscripts/mdpi/reviews/review1/changelog.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 Ms. Chen

Thanks for coordinating the submission and kindly granting us an extension for handing in our revisions. Please find a summary of the changelog below, in addition to the individual reviewer replies.

On behalf of the authors Patrick Schratz

Changelog

• Corrected an error in the code which lead NRI and VI indices use the first four corrupted bands of the hyperspectral data. This change lead to small, non-significant changes in the performance of the benchmark settings involving these feature sets.
• Moved l.55 - l.62 to the end of the introduction as suggested by Reviewer2
• Add more introductory content for the "tree defoliation" topic as suggested by Reviewer2
• Increased number of digits from two to three in Table 3, Table 4 and Table 5 as suggested by Reviewer3.
• Figure 3: Removed erroneous statement about SE values and increased digits from two to three to highlight small differences in the results as suggested by Reviewer1 and Reviewer3.
• Rephrased the introduction and reasoning on why filters methods were chosen over wrapper methods as suggested by Reviewer3
• Emphasize feature set differences earlier in the manuscript to underline the benchmarking idea of the study as suggested by Reviewer3
• Explain in more detail why a pairwise-correlation variable selection was applied as a preprocessing step as suggested by Reviewer3
• Add reference https://arxiv.org/abs/2104.04295 underlining the mentioned current research with respect to the development of robust methods for unbiased estimation of feature importance for highly correlated variables.
• Add additional references introducing other studies dealing with hyperspectral data, defoliation and tree health as suggested by Reviewer2
• Add section "Practical implications on defoliation and tree health mapping" to address the effects of the results and the response variable more directly, originated from a comment of Reviewer2
• Applied extensive language and grammar editing using professional third-party software.

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