Data Fusion Enhanced High-Dimensional SERS Fingerprints Construction via Dual-Wavelength and Multisubstrate Strategy...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Traditional label-free surface-enhanced Raman spectroscopy (SERS) provides fingerprint information for analyte identification.
• One-dimensional spectral data from conventional SERS is insufficient for complex samples like wastewater or pollution source tracing.

Purpose of the Study:

• To develop a data-fusion-driven machine learning approach for precise and robust wastewater identification.
• To create a holistic SERS fingerprint by integrating dual-wavelength and multisubstrate data.

Main Methods:

• Collected 12,000 spectra using four noble metal nanoparticle types under two excitation wavelengths.
• Employed a hybrid feature-decision fusion strategy to combine spectral features.
• Utilized optimized machine learning models and probability-level fusion for data consolidation.

Main Results:

• Achieved up to 99.67% identification accuracy for wastewater samples.
• Maintained 96.67% accuracy during blind sample testing.
• Demonstrated the generation of high-dimensional fingerprints through data fusion.

Conclusions:

• The SERSynergy approach offers efficient and accurate wastewater identification.
• This method shows potential for precise spectral feature acquisition and identity discrimination in complex matrices.