Page 55 - Read Online

Page 14 of 18 Li et al. J. Mater. Inf. 2025, 5, 29 https://dx.doi.org/10.20517/jmi.2024.103

Figure 4. Performance of six regression methods on the test and training sets for Shore hardness.

result, the overall formula approximates a positive correlation with (ω ) and a negative correlation with
4
XLD, which aligns well with chemical prior knowledge. However, in the MLR method, the coefficient for
the directly measured plasticizer content is negative, which contradicts the chemical prior knowledge.

Based on this analysis, the regression formula derived from the DSO method for the aging experimental
data holds practical chemical significance.

CONCLUSIONS
In this study, we have proposed a comprehensive evaluation framework for SR algorithms and
demonstrated its application to the aging process of rubber materials. The primary goal was to explore and
quantify the relationship between the microscopic characteristics and macroscopic properties of rubber
materials during aging, leveraging the interpretability and flexibility of SR methods. Through rigorous
validation and application of SR on experimental aging data, we have made several key findings that
contribute to the understanding of material aging mechanisms and the potential of SR in material science.

We introduced a novel evaluation framework tailored to real-world experimental data, addressing
challenges such as data sparsity, noise, and extraneous variables. Unlike traditional evaluation methods that
rely on artificial datasets, our framework provides a more accurate and comprehensive assessment of SR
algorithms, highlighting their practical applicability in complex, real-world scenarios. Our evaluation
framework identified the SR methods that performed better on the experimental dataset. This framework
ensures that the chosen SR methods are capable of uncovering meaningful relationships in material aging
data, ultimately improving their usability in scientific research.