DPMS: Data-Driven Promotional Management System of Universities Using Deep Learning on Social Media

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear authors! presented article has scientific practical significance. However, I recommend several provisions that may improve the content of the article

It is important to clearly define the objectives of the search and the main tasks

 in the theoretical review I would like to see what are the modern approaches to the formation of a strategy for the development of advertising on the basis of bighdat, How are the approaches to the collection of information formed?

The authors point out that the data were collected by stakeholders. How were these stakeholders sampled? By what method ? By what method is the number of these participants representative?

In the methodology of the study we would like to see the author describe the main stages of the study

 Finally, I would like to see the concrete results of the theoretical and practical part of the study

 I would like to see in the discussion how the result differs from the results, which are based on manual information gathering 

We would like to see the authors show the practice of application of the presented results for the enterprise

Author Response

Reviewer Comment 1:

It is important to clearly define the objectives of the search and the main tasks

Author Response to Comment 1:

We are truly grateful to the esteemed reviewer for this valuable comment. After this comment, we analyzed the article again and found the objective unclear. We modified the article by clearly defining the objective in the abstract and introduction. The core objective of this paper is to develop a decision support system to select appropriate content for social media marketing. After clearly defining it, the quality of the paper has significantly improved.

We thank the esteemed reviewer for this valuable contribution that improved the quality of the paper significantly. The modification has been highlighted in the revised manuscript.

Reviewer Comment 2:

 in the theoretical review I would like to see what are the modern approaches to the formation of a strategy for the development of advertising on the basis of bighdat, How are the approaches to the collection of information formed?

Author Response to Comment 2:

We would like to express our gratitude to the knowledgeable reviewer for this insightful comment. After receiving this comment, we further conducted a literature review to reanalyze the modern approaches. However, according to our review, none of the recent methodologies is similar to the proposed paper.

1. S. Arasu et al. [1] developed a Machine Learning (ML)-based approach to enhance social media marketing. It predicts the buyer's purchasing decision, which helps the seller to process orders faster. E. Kongar et al. [2] applied ML to target customers through social media post-mining. A consumer behavioral approach-based method developed by P. Ebrahimi et al. [3] samples the customer according to their needs, which helps sellers to maintain products in demand. Another similar research was conducted by K. Chaudhary et al. [4], where ML has been used to analyze consumer behavior. A reinforcement learning-based approach developed by P. Eklund et al. [5] analyzes the effect of advertisements. The review of the recent literature ensures that the proposed DPMS is unique and none of the concurrent approaches is similar.

We've added a new paragraph in the revised manuscript and highlighted it. After adding this paragraph, the literature review has become more robust. Thank you for this valuable comment. It helped improve the quality of the paper.

References:

1. Arasu, B.S.; Seelan, B.J.B.; Thamaraiselvan, N. A machine learning-based approach to enhancing social media marketing. Computers & Electrical Engineering 2020, 86, 106723.
2. Kongar, E.; Adebayo, O. Impact of social media marketing on business performance: A hybrid performance measurement approach using data analytics and machine learning. IEEE Engineering Management Review 2021, 49, 133–147.
3. Ebrahimi, P.; Basirat, M.; Yousefi, A.; Nekmahmud, M.; Gholampour, A.; Fekete-Farkas, M. Social networks marketing and consumer purchase behavior: the combination of SEM and unsupervised machine learning approaches. Big Data and Cognitive Computing 2022, 6, 35.
4. Chaudhary, K.; Alam, M.; Al-Rakhami, M.S.; Gumaei, A. Machine learning-based mathematical modelling for prediction of social media consumer behavior using big data analytics. Journal of Big Data 2021, 8, 1–20.
5. Eklund, P. Reinforcement Learning in Social Media Marketing. In Research Anthology on Applying Social Networking Strategies to Classrooms and Libraries; IGI Global, 2023; pp. 836–853.

Reviewer Comment 3:

The authors point out that the data were collected by stakeholders. How were these stakeholders sampled? By what method ? By what method is the number of these participants representative?

Author Response to Comment 3:

This is an excellent question. Thank you for raising it.

We followed a little different approach to data collecting, deviating a little from the traditional approach. The purpose of the proposed DPMS is to have a real impact on social media marketing. Unlike most of the papers, we didn't aim for better classification only. We aimed to increase the social media marketing impression. That is why we didn't sample the stakeholders while collecting data.

The stakeholders provided their feedback in a natural pattern. That is why it took around a year to collect enough data to develop the proposed DPMS. However, after collecting the data, the research team went through the feedback. In this case, human intelligence was used to manually label the data into four classes: Good, Average, Neutral, and Poor.

There are 22,000 active students in the experimenting university. We didn't disclose the number in the paper for the confidentiality issue. These students and their guardians are the stakeholders. The total population is around 85,000. In 12 months, we received 9461 feedback. And after dataset cleaning, there were 8810 instances. These instances where labeled with four class names manually.

We again thank the reviewer for raising this question.

Reviewer Comment 4:

In the methodology of the study we would like to see the author describe the main stages of the study

Author Response to Comment 4:

Please accept our heartiest gratitude for this excellent comment. After your comment, we've revised the paper, and it seems the paper confuses the readers without highlighting the main stages of the study in the methodology. We've explained the main stages and also created a new figure to graphically present these stages. After this modification, the quality of the paper improved noticeably.

We thank the esteemed reviewer for improving the quality of our paper by making this valuable comment. The modification according to this comment has been highlighted in the revised manuscript.

Reviewer Comment 5:

 Finally, I would like to see the concrete results of the theoretical and practical part of the study

I would like to see in the discussion how the result differs from the results, which are based on manual information gathering

We would like to see the authors show the practice of application of the presented results for the enterprise

Author Response to Comment 5:

Thank you for this remarkable comment. We appreciate the esteemed reviewer's effort to help us improve our paper's quality.

We presented the results of the proposed DPMS in two sections. They are:

1. Section 4: Experimental Results & Evaluation
2. Section 5: DPMS Implementation & Analysis

Section 4 evaluates the performance of the BiLSTM network in terms of the theoretical formula of precision, recall, specificity, and F1 score. It uses the confusion matrix concept to find the values of TP, TP, TN, and FP. An AUC-ROC analysis has also been performed to investigate the effectiveness of the performance in section 4. The results presented from these analyses are the theoretical evaluation of the proposed DPMS.

On the other hand, section 5 presents the results obtained from the practical implementation of the system. After explaining the implementation overview, the performance of human observations and DPMS's predictions have been compared. The real-life social media response has also been tracked and presented in this section. These results are the practical results of the study.

The difference between the prediction from the DPMS and manual observation has been listed in table 4 of the paper. It has been further illustrated in Figure 7. The practical implementation in real-world applications has been illustrated in Figure 6 with an explanation.

These sections have been highlighted in the revised manuscript.

Thank you, respected reviewer, for spending your valuable time reading our manuscript, providing practical suggestions, and helping us improve the paper's quality. Because of your contribution, the paper quality has improved significantly. And we are wholeheartedly grateful to you for your contribution.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript adopted an interesting problem but has failed to frame it aptly. The following are the shortcomings:-

1. the introduction section is poorly crafted and lacks information about:- what is the problem.

why that problem at hand is important to tackle?

what can be the potential solution (yes deep learning-based biLSTM) but why only the BiLSTM can resolve the problem at hand?

what are the properties or advantages of using BiLSTM over other ML/DL/TL-based techniques?

2. Agian the literature review is poorly crafted and lacks the following information:- 

the literature review should show all possible literature available to university promotions, sentiments, and deep learning.

the literature review should include information for each selected paper. such information is:- their aim, used techniques (yes you have it in paper), datasets, data features, their model's accuracies, recall, and all and their findings.

at the end of the LR you state the gaps in literature and how you are filling them.

I hope this will help you to prepare the manuscript more intuitively for future publications.

Comments on the Quality of English Language

Please proof read the paper many times and remove any self-citations or irrelevant citations.

Author Response

Reviewer Comment 1:

The introduction section is poorly crafted and lacks information about:- what is the problem.

why that problem at hand is important to tackle?

what can be the potential solution (yes deep learning-based biLSTM) but why only the BiLSTM can resolve the problem at hand?

what are the properties or advantages of using BiLSTM over other ML/DL/TL-based techniques?

Author Response to Comment 1:

We are delighted to receive these valuable comments from the respected reviewer. Please accept our heartiest gratitude for your valuable contribution.

After reading your comments, we critically analyzed the introduction. We also felt the same that the problem statement is not well defined. As a result, the paper suffered from quality issues. However, we've rewritten the first two paragraphs of the introduction. This time, we focused on the problem statement. Then, we expanded the introduction, mentioning why the problem needs to be addressed.

In the second paragraph, we briefly mentioned why the BiLSTM network was used. We understand that the paper is incomplete if we don't mention why BiLSTM has been used and what are its properties and advantages. To answer these questions, we briefly expressed the reason for using the BiLSTM network in the introduction. Later, we explained it in detail in Section 3.3. The title of this section is 'Model Selection and Architecture.' Here, we analyzed the features of the BiLSTM network rigorously and presented its logic in this paper.

All modifications have been highlighted in the revised manuscript. We again thank the esteemed reviewer for significantly improving the quality of the paper. We wish we could personally thank this reviewer face to face for his brilliant observation and for guiding us to improve the paper's quality significantly.

Reviewer Comment 2:

Agian the literature review is poorly crafted and lacks the following information:- 

the literature review should show all possible literature available to university promotions, sentiments, and deep learning.

the literature review should include information for each selected paper. such information is:- their aim, used techniques (yes you have it in paper), datasets, data features, their model's accuracies, recall, and all and their findings.

at the end of the LR you state the gaps in literature and how you are filling them.

I hope this will help you to prepare the manuscript more intuitively for future publications.

Author Response to Comment 2:

We are wholeheartedly grateful to this esteemed reviewer for making these valuable, insightful, and well-defined comments. These comments helped us to convert our paper into a high-quality paper.

After receiving this comment about the weakness in the literature review, we followed the instructions of this reviewer. We reviewed papers related to sentiment classification using Deep Learning and Machine Learning. We added the accuracy of the models and summarized the comparable papers in a table. Finally, a subsection titled 'Research Gap Analysis' has been created where the gaps in existing research have been discussed.

The modifications have been highlighted in the revised manuscript. After this modification, the paper's quality has been significantly improved. Now, it seems like a standard paper. We again thank this esteemed reviewer for guiding us to improve the quality of the paper.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have revised their work and tried to address all review comments.

Comments on the Quality of English Language

Minor spellchecks and proof reading are needed.