Bull Breeding Soundness Assessment Using Artificial Neural Network-Based Predictive Models

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Comments on the article titled: “Bull Breeding Soundness Assessment using Artificial Neural Network-Based Predictive Model” by Luis Marín et al.

 

General comments:

The work is very interesting and indicates that traditional BBSE methodologies have limitations in terms of accuracy and efficiency. They are subjective, time-consuming, and may not accurately predict bull fertility. Additionally, the authors note that traditional BBSE methods may not be able to identify bulls with subclinical reproductive disorders, which can negatively impact herd fertility.

Authors suggest that the artificial neural network-based predictive model has several potential benefits for bull breeding soundness assessment. Specifically, the authors note that this approach is more accurate and efficient than traditional BBSE methods, can identify bulls with subclinical reproductive disorders, and can help improve herd fertility. Additionally, the authors note that this approach can be used to develop customized breeding programs that are tailored to the specific needs of individual herds. Overall, the authors suggest that this approach has the potential to revolutionize the way that bull breeding soundness is assessed and managed.

 

Materials and Methods:

An artificial neural network-based predictive model that uses machine learning algorithms has been used to evaluate bull breeding soundness. The model used a variety of variables, including genetic group race, body condition score, age, scrotal circumference, semen volume, sperm concentration, individual sperm motility, gross motility category, colour, density, libido, pregnancy rate, cows, and calving interval.

The model was trained on a dataset of bull breeding soundness evaluations and used this data to predict the likelihood of a bull being classified as "satisfying," "unsatisfying," or "bad" . The authors note that this approach overcomes many of the limitations of traditional BBSE methods and can accurately predict bull fertility.

However it is very difficult to understand how the created model works as the details have not been given explicitly and the whole article is not structured in a satisfactory manner.

Please pay attention to the following issues:

1.                  In neural classification, there are several important factors that affect the effectiveness of the model. Here are some key considerations:

Training data quality: The training data must be comprehensive, accurate and representative of the different classes. The better the quality of the data, the better the performance of the model. The Authors try to show some problems as unequal initial data distribution (Fig 1). Is Fig 1 correlated to Table 2 ? It is very hard to follow the Author’s ideas how the most representative variables were choosen.

2.    Can U explain some inconsistency in the text regarding whether the classification is multi-class or binary. In page 9, the K-means algorithm was used to find two clusters, and only two classes distinguish clearly between both two groups. The figure 5 shows a dichotomous (binary) classification: Accepted for class A, and Rejected for B and C. However, in Page 10, the confusion matrix shows three classes: Class A, Class B, and Class C. The accuracy of the clustering analysis with the K-means algorithm was 78%. When using this approach to construct the third class, the separation of the C class remained unclear, hence the problem was reduced to a binary classification comparing the A class to the B and C classes. So, it is unclear whether the classification is multi-class or binary.

3.    Class balance: For multi-class classification, it is important to have an equal number of samples for each class so that the model does not favour one class over others – this condition is not met.

4.    There is no explicit model architecture described in materials and methods. The choice of the appropriate type of model (e.g. convolutional neural network, recurrent neural network) depends on the type of data and the classification problem. The appropriate architecture of a neural network can significantly affect its effectiveness. Number of layers and neurons is given only in Results section, which should contain only information on the best results.

5.    Learning process, data partitioning: It is valuable to divide the data appropriately into training, validation and test sets to assess the effectiveness of the model. There is no information given about this issue.

 

6.    Please pay the attention to the formatting of an article

a)    Working place should be listed one time, if more authors from one institution please add additional e-mail details as follows:

¹ School of Electrical and Electronic Engineering. Universidad Veracruzana, Veracruz, Mexico; [email protected] ; [email protected]

b)    Please format Table 2 to the standards of Agriculture journal

c)    Ethical statement should be placed at the end of the article, after Funding

d)    Please pay attention to a consecutive page numbering

e)    Please check the formation of the references. Some references are written with different font and some do not comply with the requirements of Agriculture journal

 

 

Comments on the Quality of English Language

English mostly correct. Few sentences require re-writting (such as lines 57-58, 75-80, 197-203)

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

In this paper, authors well described an artificial neural network-based predictive model for bull breeding soundness evaluation, which is easy to understand. Here are some questions that the authors have to address to complete this paper.

1.Line 129: morphological and behavioral evaluation, the variable measured methods and evaluation standards were before 2010, what is your consideration? Do you have any latest methods?

2.Lines 179-182: all the bulls fit into one of the following categories: A class, B class, C class. Please provide a volume value or a detailed introduction.

3.Line300, Line 313, Line329: Figure 6-8, the horizontal axis is not very clear.

4. Line329: Test for ANNs with two hidden layers and different hyperparameters trained with only five input variables. What are the five parameters and why? Please explain clearly.

5.Line 398: compared to 81% for traditional BBSE done by local farmers, using only the physiological and morphological features of the animal. We cannot conclude from the paragraph with Lines 183-186 that 81% was done by local farmers. Please explain clearly.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Please see the attachment

Author Response File: Author Response.pdf