Spectrum Quantitative Analysis Based on Bootstrap-SVM Model with Small Sample Set.

Area of Science:

• Analytical Chemistry
• Materials Science
• Machine Learning

Background:

• Accurate prediction of bitumen penetration index is crucial for asphalt mixture performance.
• Traditional spectrum quantitative analysis methods struggle with small sample sets.
• Developing robust models for limited data is an ongoing challenge in materials analysis.

Purpose of the Study:

• To propose a new spectrum quantitative analysis method using a Bootstrap-Support Vector Machine (SVM) model for small sample sets.
• To build and validate a model for predicting the bitumen penetration index.
• To compare the proposed method's performance against established techniques like Partial Least Squares (PLS) and standard SVM.

Main Methods:

• Collected 29 bitumen samples from 6 companies for model development.
• Developed a spectrum quantitative analysis model utilizing the Bootstrap-SVM approach.
• Conducted comparative experiments predicting bitumen penetration index using Bootstrap-SVM, PLS, and SVM.

Main Results:

• The proposed Bootstrap-SVM model achieved the minimum prediction root mean squared error (RMSE).
• Experimental results validated the feasibility and effectiveness of the Bootstrap-SVM method.
• The model demonstrated superior performance in predicting bitumen penetration index with a small sample set.

Conclusions:

• The Bootstrap-SVM model offers a robust and effective solution for spectrum quantitative analysis with limited data.
• This approach provides a novel pathway for building predictive models when sample availability is constrained.
• The method successfully addresses the challenge of small sample sets in quantitative spectral analysis.