An Assemble Based on Clustering and Monte Carlo for the Wavelengths Selection of Excitation Emission Fluorescence Spectra

Excitation-emission fluorescence spectra is very effective to predict the concentration of organics in samples. However, redundant information and noises in the excitation-emission matrix (EEM) decrease the accuracy of the prediction concentration. Here we proposed a method to select more useful excitation and emission spectra from the EEM to increase the accuracy of prediction concentration and reduce the processing time. First, the excitation wavelengths were selected based on the clustering method to limit the redundant information in the EEM. Then the emission wavelengths were selected based on the Monte-Carlo method. To validate this method, we established the concentration prediction model with the spectra corresponding to the selected wavelengths by partial least square regression and predicted the multi component concentrations in the test samples. Our studies indicate that incorporation of this method increases the accuracy of the prediction concentration of organics and reduces the processing time.