We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Identification of marine microplastics by a combined method of principal component analysis and random forest for fluorescence spectrum processing
Summary
Researchers developed a combined principal component analysis and random forest method to identify microplastics from overlapping fluorescence spectra. The technique achieved 99.7% accuracy for component identification and a correlation coefficient exceeding 0.99 for predicting microplastic concentrations. The model, initially trained on commercial plastic samples, was also successfully applied to identify real marine microplastics.
The severely overlapped laser-induced fluorescence spectra between different microplastics pose significant challenges on fluorescence-based particle identification and quantification. To address this problem, this paper proposes a combined method of principal component analysis (PCA) and random forest (RF) for fluorescence spectrum processing. The key idea is to identify the overlapped PCA scores of the first three principal components of fluorescence spectra by the random forest method. Both pure and mixed microplastics samples were used to verify the accuracy of this method. It was demonstrated that both the compositions of the samples and mass concentration of one specific microplastics can be accurately identified. The accuracy for component identification reaches 99.7 % and the correlation coefficient between the predicted and actual concentration exceeds 0.99. Furthermore, the PCA-RF model established with commercial plastic samples was also applied for real marine microplastics identification with good identification results obtained.