An Adaptive Machine Learning Approach to Sustainable Traffic Planning: High-Fidelity Pattern Recognition in Smart Transportation Systems

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript is very well-structured and technically sound. The introduction provides solid context, the methodology is detailed and reproducible, and the results are clearly presented with strong statistical validation. Conclusions are fully supported by the findings.
Minor suggestions:

• While the English is already good, some long sentences could be shortened for even greater clarity.
• Consider expanding slightly on the real-world applicability in future research directions, especially regarding computational efficiency in large-scale deployments.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This is a valuable study for smart transportation sustainable planning, proposing an adaptive cascade clustering approach that integrates HDBSCAN and k-means via a data-driven weighted voting mechanism. It validates the model using a SUMO-simulated transport network of Khmelnytskyi, Ukraine, achieving promising results and highlighting its potential for traffic pattern recognition. However, improvements are needed in method detail, experimental comprehensiveness, and practical relevance. The following revisions are suggested:

1. It is suggested that the manuscript be polished by professional technical English editors. Focus on refining the introduction, algorithm description, and experimental analysis to ensure accurate terminology and clear logic, so readers can easily grasp the research background, technical approach, and key conclusions.
2. In Section 2.6.1 , the paper sets the weighting factors (α,β,γ) in the quality metrics for HDBSCAN and k-means to 1/3 each but provides no justification for this choice. There is also no analysis of how weight adjustments affect the voting result and final clustering quality. It is suggested to supplement a sensitivity test and report how different weight combinations influence key metrics, to verify the rationality and stability of the weight setting.
3. In Section 3, the study relies solely on a calibrated SUMO simulation dataset (RMSPE < 15%) for validation, but simulated data presets traffic scenarios and lacks the complexity of real-world traffic. This limits the model’s demonstrated generalizability. It is recommended to supplement validation using real traffic data and compare performance differences between simulated and real datasets, to confirm the model’s applicability in practical scenarios.
4. In Section 3.3, robustness tests only introduce Gaussian noise to the data, but real traffic data often contains non-Gaussian anomalies. The paper does not evaluate the model’s performance under these more realistic anomalies. It is suggested to add robustness tests for typical traffic anomalies and report changes in metrics such as ARI and temporal coherence, to better reflect the model’s practical resilience.
5. In Section 4.4, While the paper notes HDBSCAN’s quadratic complexity (O(K²)), it only provides runtime for the current dataset without analyzing scalability—for example, how runtime increases with more time windows or more intersection nodes. This is critical for real-time traffic management applications. It is recommended to add a scalability experiment, plot runtime vs. data scale curves, and discuss optimization directions for large-scale data.
6. In order to improve the article, several papers are worthy of reference:

[a]A Resilience Recovery Method for Complex Traffic Network Security Based on Trend Forecasting[J]. International Journal of Intelligent Systems, 2025, 2025(1): 3715086.

In summary, this study has its own contributions, but improvements are needed in terms of presentation and the issues mentioned above.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

All the problems are solved.