Multi-Dimensional Feature Fusion and Enhanced Attention Streaming Movie Prediction Algorithm
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors The primary research question addressed is How can we improve the short-term forecasting accuracy of streaming movie performance, given the challenges of complex, multi-dimensional, and limited datasets. The authors propose a novel prediction framework, FFLEA, which integrates multi-dimensional feature fusion, attention mechanisms, and data augmentation to enhance model robustness and accuracy. The topic is both original and highly relevant. It targets a growing area in the intersection of AI and digital media—streaming movie analytics. The paper directly addresses several existing gaps: incomplete feature representation, poor adaptability to limited data scenarios, lack of dynamic attention mechanisms in time series movie prediction. By proposing a more comprehensive and context-aware model, the paper contributes to filling these gaps effectively. Compared to previous work, this research offers several unique contributions: Fusion of multiple feature types into a unified LSTM-attention pipeline, Enhanced attention mechanism that dynamically weighs temporal and feature importance. , Data augmentation techniques (flipping and scaling) tailored to the box office prediction domain—helping generalization with limited data.This approach pushes the boundary of what has been done in movie performance forecasting. Some areas for methodological enhancement include:External validation on data from other platforms (e.g., Netflix) to test. The conclusions are consistent with the evidence and appropriately address the research question. The authors support their claims using extensive experiments.Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThe article “Multi-Dimensional Feature Fusion and Enhanced Attention Streaming Movie Prediction Algorithm” proposed and applied the FFLEA algorithm for feature extraction and solving issues of finiteness, data characteristics, and influencing variables in streaming movie prediction to support business decisions. The article is interesting and novel since it uses diverse Machine and Deep Learning models combined to solve the feature extraction model. The FFLEA model overcomes other Machine Learning approaches. Although the paper is relevant, I have found diverse issues the authors need to clarify.
1.- The document repetitively stated that is required to extract film or movie features; however, it is not clear what those are. I suggest elaborating a scheme or chart listing the main features to detect by the FFLEA.
2- It is not clear what the acronym FFLEA stands for, Feature Fusion and Enhanced Attention, what the “L” stands for? long- and short-term memory (LSTM)? Then the model should be FFLSTMEA?
3.- What software, programming language was used to build the Machine Learning algoritm?
4.- Results section: diverse Machine Learning methods such as CNN, LSTM, BiLSTM, GRU, Transformer, CNN-Transformer, 517 TDformer, MTPNet, SageFormer, Crossformer, and FFLEA were used for feature extractions, and these were validated with MAE, RMSE, MSE, R2, … metrics. However, it is not clear what film or movie feature is evaluated?
5.- The analysis of projected results or discussion, in my opinion, was not properly performed. It should include a comparison with results reported in the scientific literature, mentioning the limitations and advantages of the proposed algorithm.
6.- What would be the next or future applications of FFLEA?
7.- Read carefully the document to detect typos
Line 66: ueses
Correct Tabel by table, (table 1 and 2) line 524, 558.
8.- Figure 8 quality and size need to be improved, the label can hardly be read within the figure frame.
Comments on the Quality of English LanguageEnglish is fine, although authors need to detect diverse typos in the document
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors of the paper introduced a so-called fusion long- and short-term memory enhanced attention network (FFLEA) model for streaming movie prediction purposes. The characteristics of the proposed model were partly verified experimentally. The content of the article meets with the topics of Applied Sciences journal.
The English grammar of the article must be checked once again (e.g., “complex market reality.market realities”; “forecasting[23]” – many times, whitespace between the words is missing; word “data” has only plural form; etc.).
In general, the topic of the article is interesting. However, several parts must be better presented. The article must be reworked and extended. The article needs more than major revision.
Notes:
- Abstract / Paper – all the used abbreviations in the paper abstract (and in the paper text) must be clearly defined.
- Introduction / Section 2 – the state-of-the-art (SoA) is evaluated on an acceptable level. However, the main contributions of the work (and more in detail) should be presented after the elaboration of the related works.
- Introduction / Section 2 – the main differences between the actual papers and authors‘ previous papers should be clearly explained.
- Introduction – information about the organization of the paper is missing.
- Section 3 – the visual quality of the Fig. 1 must be improved.
- Section 3 – the architecture of the proposed concept must be presented in detail (Fig. 1 in a more detailed form) and discussed in detail.
- Article – sometimes, it is not clear that which equations have been derived by the authors and which ones are used from other works.
- Section 4.1 – “The data for this study comes from the iQIYI streaming movie list applet…” – link on appropriate sources is missing.
- Article – Figs. 3-5 and Fig. 8 are unreadable – bad visual quality, small sizes, blurred pictures.
- Section 4 – all the used HW/SW equipment should be presented in detail (please, check it once again). Next, for reproducible research (for its support), I would suggest the authors, if possible, make all the complete mathematical model publicly available.
- Section 4 – Fig. 6 and Table 1 – in my humble, both contain the same data. Hence, redundant information must be reduced. Next, the results obtained are not discussed and analyzed in detail. Why are some rows in Table 1 highlighted in yellow?
- Section 4 – according to the obtained graphs, the proposed architecture has comparable performances in terms of some SOTA solutions. Please, discuss it in detail.
- Section 4 / Conclusion – it is not clear how can the proposed concept applied in real-world (real-time) scenarios. Such a discussion (in detail) is missing.
- References – in my humble opinion, this part of the paper does not respect the journal‘s template.
The English grammar of the article must be checked once again (e.g., “complex market reality.market realities”; “forecasting[23]” – many times, whitespace between the words is missing; word “data” has only plural form; etc.).
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsAll my concerns have been addressed by the authors,
Author Response
Thank you very much for reviewing my manuscript and for recognizing it, and I will follow up with more progress.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe article has been partly improved. Many thanks for the explanation letter! Next time, please, also indicate in the explanation letter that where the changes provided can be found in the paper. Thank you!
I have the following notes:
- Abstract – please, reduce the length of the paper abstract.
- “Introduction / Section 2 – the main differences between the actual papers and authors‘ previous papers should be clearly explained.” – I mean the authors of the actual paper and no other authors of other papers.
- Section 4.1 – the presented link “#小程序://网络电影榜/nWbRg2s87h8Shxg” is invalid. Please, check it.
- Section 4 – what is the main purpose of Fig. 5? Please, explain it.
- Section 4 – please, remove the rectangular around Figs. 3 and 4.
- Section 4 – it should be “Python 3.7”. Next, there is another typo “As Tabel…”.
- Section 4 – in my humble opinion, subfigures in Fig. 6 should be replaced by bar graphs. What is the difference between Table 1 and Fig. 6?
- Section 4 – for reproducible research (for its support), I would suggest the authors, make all the complete mathematical model publicly available.
- Section 4 – In my humble opinion, the authors should consider and use more advanced and comprehensive objective metrics and statistical analysis. For inspiration, it is possible to check the following paper: “Performance Evaluation of Objective Quality Assessment Methods for Omnidirectional Images Under Emerging Compressions”.
- Article - the English grammar of the paper must be checked once again (see some comments above).
See "Comments and Suggestions for Authors".
Author Response
Please see the attachment.
Author Response File: Author Response.docx