An Adaptive Holt–Winters Model for Seasonal Forecasting of Internet of Things (IoT) Data Streams

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1.    The manuscript lacks any form of statistical significance testing (e.g., p-values, confidence intervals), making it unclear whether the observed performance gains are truly meaningful.
2.    Despite claiming that the AHW model is computationally efficient, the paper does not provide any runtime or latency benchmarks compared to baseline models.
3.    The authors should conduct ablation studies to isolate the individual contributions of the recent window and historically similar window components.
4.    The abstract emphasizes that AHW is both precise and fast, but the "fast" claim is not empirically validated and should be supported with concrete runtime analysis.
5.    Although the model is framed as suitable for real-time applications, there is no real-time or streaming scenario used to validate this aspect of the method.
6.    The rationale for the study would benefit from deeper discussion and citation of related work using pattern similarity in time-series forecasting.
7.    The paper should discuss its limitations, especially regarding AHW’s sensitivity to seasonal assumptions and its adaptability to non-periodic data.
8.    The paper currently assumes that RMSE alone is sufficient for comparison, but additional evaluation metrics (e.g., MAE, MAPE) could enhance the reliability of findings.
9.    While the bar plots are clear and readable, the choice of colors and styling lacks a professional scientific appearance; the authors should revise the plots using a more refined color scheme and formatting to enhance visual quality.
10.    The authors are recommended to cite the paper “SmartFormer: Graph-based transformer model for energy load forecasting” (F. Saeed et al., 2025), which presents a transformer-based deep learning approach for time-series prediction. Including this reference would strengthen the related work section by providing a modern benchmark for IoT-related forecasting and justifying the computational simplicity of AHW in contrast to graph-based deep learning models.
11.   Referencing some works that support the generalizability of lightweight forecasting models like AHW and help contextualize the cross-domain applicability of time-series forecasting approaches.

Author Response

A point-by-point reply file has been uploaded 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper presents an interesting idea stemming from a concerning issue in time series prediction. There is merit in the work, as it uses TinyML concepts to implement a predictive model with higher capacity than traditional approaches. However, there are several points to improve:

1. The RMSE equation presented is not a traditional one. Detail how it was adapted in this work or provide its source.
2. While the results are competitive, the authors mention that the SOTA (state-of-the-art) uses RNNs and CNNs for time series prediction, but these models are not explored in the evaluation.
3. The authors should correlate the model's predictive capability with the underlying dataset characteristics.

Author Response

A point-by-point reply has been uploaded. 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

To the Authors,

The manuscript submitted for review, "An Adaptive Holt-Winter Based Model for Forecasting High-Volume IoT Data Streams", in IOT magazine presents a modification of the Holt-Winters algorithm. The manuscript is written in technically sound English and its sections are clear and concise, covering topics such as big data, current state of knowledge, proposed algorithm, and testing against classical methods. The manuscript is well-structured and provides necessary information to the reader without being overly verbose.

The proposed application of the algorithm is focused on IoT data; however, it seems that the algorithm could have a more general applicability to medium and large datasets that are time series in nature. It would be interesting to see how the algorithm performs with noisy environmental data.

Comments and Suggestions for the Manuscript:

1. Consider expanding the acronym in the title.
2. In the Introduction section, it would be beneficial to mention predictive methods adapted for IoT data.
3. The section lacks a detailed discussion of the modifications made to the Holt-Winters method, which would help to distinguish the proposed algorithm from existing solutions. Please add a description of the modifications to the Holt-Winters method that have appeared in the literature over the past few years.
4. When presenting the modifications to the Holt-Winters algorithm, consider adding equations to remind the reader of the classical model.
5. Figure 1: Please change the background color and increase the font size. In black and white print, the graphic is difficult to read.
6. Figure 2: The y-axis scale should start from 0.
7. In the evaluation, you focus exclusively on RMSE. Consider expanding to other metrics such as AIC, MAE, etc.

Author Response

A point-by-point reply has been uploaded

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I am delighted and grateful to the authors for their diligent and thoughtful revisions, which have markedly improved the manuscript’s clarity, consistency, and responsiveness to prior feedback.  As you continue to refine the manuscript, please focus on addressing the suggestions outlined below without introducing new content; instead, rigorously correct and polish the existing text to ensure the highest standards of scientific rigor and English language quality.

Revise the title to use “Holt–Winters” (not “Holt-Winter”) and, unless you demonstrate throughput or scalability on very large streams, remove “High-Volume” so it simply reads “IoT Data Streams.”

Recommended Title: “An Adaptive Holt–Winters Model for Seasonal Forecasting of IoT Time Series”

Perform a targeted literature check on ensemble ETS variants, nearest-neighbor forecasting, and dual-seasonal ETS; if no prior work combines RMSE-based closest-window selection with two Holt–Winters fits and adaptive blending, you may keep the novelty claim, otherwise rephrase AHW as an incremental improvement.

Rewrite both Abstract and Conclusion in concise scientific English, eliminating redundancy and ensuring a clear flow from problem statement through methodology to key results and implications.

Please clarify and standardize your dataset references in the Abstract: for example, specify “temperature data from the National Climatic Data Center (NCDC),” “humidity and soil-moisture measurements from the Basel City environmental dataset,” and “global intensity and global reactive power from the Individual Household Electric Power Consumption (IHEPC) dataset.

Please explicitly name the five datasets in the Conclusion also to reinforce the study’s empirical foundation and help readers immediately grasp the scope and applicability of your findings. 

On pages 8–9, correct the pseudocode so loop counters (e.g., i, j, k) advance and terminate properly—remove any manual increment inside for loops—compute the number of windows once, and present each algorithm step by step with matching loop delimiters.

Use “Holt–Winters” uniformly throughout the manuscript; avoid introducing variants like “traditional,” “conventional,” or “original.”

On page 12, combine the monthly and daily plots into one figure with panels (a) and (b) and apply this caption:

Fig. 2. Comparative evaluation of RMSE and MAE across eight forecasting methods on the NCDC temperature dataset (DB1): (a) monthly prediction; (b) daily prediction.

Remove Figures 4–11 on page 14, as they duplicate data already summarized in Table 2.

On page 15, specify the statistical test used for p-values (e.g., paired t-test on per-window RMSE), revise the discussion of Table 3 to focus strictly on statistical findings, and remove the unsupported claim in lines 472–474.

On page 16, rewrite the discussion of Table 4 as a formal scientific interpretation—instead of informal commentary.

Extensive English proofreading is highly needed to correct grammar, clarify phrasing, and ensure the manuscript meets professional scientific standards.

Author Response

A point-by-point reply has been attached. 

Author Response File: Author Response.pdf