Shoplifting Detection Using Hybrid Neural Network CNN-BiLSMT and Development of Benchmark Dataset

Round 1

Reviewer 1 Report

The manuscript titled “Shoplifting Detection using Hybrid Neural Network CNN-BiLSMT and Development of Benchmark Dataset” is well written and described. In this paper, authors attempted to develop a benchmark dataset for shoplifting detection. The developed dataset validated on the proposed model and yields acceptable performance. The readability of the manuscript is further improved by the incorporation of following key observations:

1.       Authors must include the reason behind the selection of Inception and BILSTM model to develop hybrid system.

2.       The author must include the reason behind the selection of BILSTM model to predict shoplifting.

3.       There are so many typos error and grammatical mistakes in the manuscript, so have to carefully read the manuscript and do the required corrections.

4.       The proposed model is validated on the mentioned dataset using in terms of accuracy only. According to reviwer’s opinion, accuracy is not the sufficient parameter to decide the stability of the system. So, the author must include more qualitative analysis to determine the performance of the proposed model.

5.       In the result section, Table 2 shows the accuracy for relu activation function only. What will be the behavior with other activation function with respect to the developed algo.

6.       The author needs to describe more details about the result & discussion.

7.       Only mentioned accuracy is not suitable, so the author has to include ROC curve, and confusion matrix also.

8.       The used reference is not in the proper order and format. So do the needful corrections and include latest reference also.

Minor corrections required.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

In my opinion. the article needs to be improved for the following reasons:

(1)The last sentence of the abstract seems unfinished and needs to be improved.

(2) p.6 Table 1, What is the meaning of 4,60,800?

(3) p.10  I think other major indicators of machine learning, such as specificity, f1 score, AUC... needs to be added.

(4) I think the ethical issue needs to be mentioned. Whether it is suitable for researchers ask other persons to mimic the criminal behavior in order to gather the related information is doubtful.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The idea behind the study is interesting, but I have some concerns that need clarification:

To understand the behavior of the CNN model and its ability to generalize, it is important to analyze the error curves, comparing the performance on the training and validation sets.

Accuracy alone does not consider class imbalance, systematic errors, or other nuances of the dataset. Relying solely on accuracy as a metric to evaluate model performance may overlook these factors. It is important to consider additional metrics that provide a more comprehensive assessment of the model, such as precision, recall, F1-score, or area under the ROC curve (AUC-ROC).

Regarding the optimization of the deep learning ensemble model, it would be helpful to know the specific techniques or approaches used in the optimization process. Additionally, explaining the rationale behind choosing a particular architecture for the model would provide further insights into the decision-making process.

The 80/20 split for dataset partitioning is a common practice, where 80% of the data is used for training and the remaining 20% for validation or testing. However, it would be beneficial to clarify whether the split was made randomly or if any specific considerations were taken into account during the partitioning process.

The section labeled as "Results and Discussion" appears to be more suitable as a conclusion section, and it would be appropriate to make this correction to ensure the coherence of the report's structure.

In Line 394, a reference is missing. Please provide the missing reference or revise the sentence accordingly.

Some figures need to be improve in their resolution and quality.

The paper is readable but the english can be improve.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The authors clarified all my concerns and questions. The paper is ready to be published.