A Machine Learning Approach to Predict Customer Usage of a Home Workout Platform

Round 1

Reviewer 1 Report

The authors propose a machine learning approach to assess the activity
of users on a home workout platform. The authors quantitatively
evaluate the factors that keep people using a home workout app by
applying SVM, Random Forest, K-nearest neighbor, and logistic
regression. The data was crawled from "Keep," a mobile app available
in the market. They implemented the data crawler for collecting data
from Keep. They found that the random forest effectively predicted the
continuous usage of users after installing the home workout
application. Also, they have found that certification, which means
verification of user's identity through peer-review, is the most
critical factor for the continuous workout. This indicates that
providing factual information verification, such as ID card
information, mobile phone verification, or email, highly motivates
users to continue the workout.

The paper is objective and carefully written and will be valuable to
readers of this journal.

The reviewer has one concern about privacy data automatically clawed
by authors.  They used the training set data of 7,734 Keep users, but
the reviewer could not find agreement from those users. If the authors
did not receive consent forms from users, the experimental results
could not be published. The authors must explain whether users agreed to these experiments or not.

Some small revision:
- "logic regression" at P.1 L.12 must be "logistic regression".

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1- Where is the novelty in the paper? 

2- The machine learning approach is not described properly. It can be represented by a model or an algorithm. 

3- If there is a code it should be printed.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The works represent a straightforward toy example ML.

The greatest limitations are:

1) The abstract should present the background, the research issue, a quick summary of experiments/results and a take-home message (i.e., conclusion). At this point, it is not clear what platform and Keep are.

2) COVID 19 and ML contexts are too generally presented. My suggestion is to search and present specific works that are closely related to the tackled issue.

3) The code of the crawler, as well as the code of the analysis, should be in a git with public access, not as a print screen.

4) It is not clear the target class from ML point of view.

The overall approach looks like a simple sun of ML algorithms with very few novelties.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

All the problems pointed out in the 1st round review were cleared.