Machine Learning and Deep Learning Models for Dengue Diagnosis Prediction: A Systematic Review

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The review is well written. Figures 4 could have been represented in a better way. 

I hope the samples mentioned could have been either signal or image. Not mentioned in the paper. Maybe not needed. 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Comments attached in the file

Comments for author File: Comments.pdf

Comments on the Quality of English Language

no

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Abstract

·      The total number of different algorithms evaluated across the 32 studies should be included. This will allow for better understanding from the boarder readership. Also a little confusing how we go from 32 studies to 48 algorithms. 

·      “The results revealed that the most outstanding algorithms in terms of performance 25% of the studies reported that Support Vector Machine (SVM) showed the most outstanding performance, followed by Random Forest with 15.62% of the 32 articles analyzed.” This sentence is difficult to understand. Please revise. 

 

Introduction

·      This section is well written and does a nice job of describing the scope and issues surrounding Dengue and it surveillance. 

·      The introduction would benefit from a slight discussion of typical symptoms observed in patients with dengue. This would also be beneficial since the authors are reviewing predictive algorithms based on early symptoms. 

·      For the instances following the first occurrence of “World Health Organization” the authors can use acronym instead of spelling out “World Health Organization” again. 

·      Purpose of the study is also clearly presented. 

Methods

·      In section 2.1 it should be made clear the dates, or date range in which searches were conducted to identify articles.

·      Please include how many investigators were used to review or extract the relevant data from the articles as this is not clear in the methods section. Further, if multiple investigators were used it should be clearly stated how any discrepancies were resolved (if any).

·      Table 3, should participates with confirmed dengue also be included?

·      A brief description of what data elements were extracted from articles and how they were categorized into domains should be included in the methods section.

·      It is not clear what the value 63.16 represents after the 79. Is this the percent of articles included after applying the inclusion and exclusion criteria, or after removing duplicate articles? This is also not easily derived from the flow chart. 

 

Results

·      What are examples of the “other reasons” articles were excluded?

·      The domains listed in section 3.2 should also be discussed in the methods section. How were these domains derived and how were articles determined relevant to each domain?

·      When discussing the number and percents of studies each region contributed, stay consistent with either listing the percent or number or studies first. 

·      How was “outstanding performance” determined?

Discussion

·      The authors should consider adding discussion on the computational load and drain should be considered for each model and the availability of these models to be performed at different localities for those trying to prevent Dengue. There is not doubt that SVM and Radom Forest are powerful predictive tools, however, not every health or preventative agency will have the computational capacity to incorporate these models into their efforts.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf