Discrimination of the geographical origin of Codonopsis pilosula using near infrared diffuse reflection spectroscopy coupled with random forests and k-nearest neighbor methods

The combination of near infrared (NIR) spectroscopy with chemometrics provides an approach to study Codonopsis pilosula according to its geographical origin. Firstly, principle component analysis (PCA) was used to group samples based on their spectral differences. Random forests (RF) and k-nearest neighbor (KNN) were applied to build the classification models and predict the geographical origins of test samples. Raw and SNV first derivative NIR spectra were compared to develop a robust classification rule. Feature selection by RF using the variable importance returned 4 selected features, and the selected effective wavenumbers were put into KNN to establish the classification model. For independent test set, same total accuracy rate 94% could be achieved using RF and KNN. These results showed that NIR combined with chemometrics might be a suitable method that can be easily implemented to classify C. pilosula.