I appreciate the authors' effort to address a relevant and impactful topic in pavement engineering through a data-driven machine learning approach. The paper is clearly structured and presents good results using Random Forest regression for rolling resistance prediction. Nevertheless, several areas require improvement and my comments are as follows: 

1. Minor grammatical errors are present throughout the manuscript. A thorough proofreading is necessary.

 2. Abstract: effectively communicates the study's purpose and findings. However, the problem statement and research objective could be articulated more clearly to emphasize the gap in current knowledge and the motivation behind using machine learning.

 3. Introduction: is well-written and clearly identifies the research gap and objective. As a minor point, please note that “greenhouse” should be written as one word, not “green house”. 

4. Methodology: please expand the acronym “RF” (Random Forest) at its first mention in the methodology section to ensure clarity for readers.

 5. Data collection: lacks sufficient detail and the authors should explicitly state the total number of data points used in model development and testing, the specific characteristics of the road network where data was collected, and whether this road network is representative of typical Dutch conditions or specific to a site or project. 

6. Sections 2.2: given the likelihood of non-linear interactions among variables, would Spearman’s rank correlation be a more appropriate choice than Pearson correlation in this context? Additionally, the manuscript should clearly list which features were excluded from the Random Forest model. 

7. It is recommended to create a separate section that lists all underlying statistical assumptions of the models used, including normality, independence, multicollinearity, and feature relevance.

 8. Parameters such as tire pressure vary across vehicle types, and ambient air temperature is highly location-specific. The authors should elaborate on how such variability was accounted for in the model and whether the current approach can generalize beyond the tested conditions. 

9. A brief discussion on the limitations of the model such as generalizability or input data availability would enhance critical evaluation. 10. Although the conclusion mentions future work, more concrete suggestions on how this work could be included to guide subsequent research efforts will be useful.