Data Fusion of Fourier Transform Mid-Infrared (MIR) and Near-Infrared (NIR) Spectroscopies to Identify Geographical Origin of Wild Paris polyphylla var. yunnanensis

Abstract

Origin traceability is important for controlling the effect of Chinese medicinal materials and Chinese patent medicines. Paris polyphylla var. yunnanensis is widely distributed and well-known all over the world. In our study, two spectroscopic techniques (Fourier transform mid-infrared (FT-MIR) and near-infrared (NIR)) were applied for the geographical origin traceability of 196 wild P. yunnanensis samples combined with low-, mid-, and high-level data fusion strategies. Partial least squares discriminant analysis (PLS-DA) and random forest (RF) were used to establish classification models. Feature variables extraction (principal component analysis—PCA) and important variables selection models (recursive feature elimination and Boruta) were applied for geographical origin traceability, while the classification ability of models with the former model is better than with the latter. FT-MIR spectra are considered to contribute more than NIR spectra. Besides, the result of high-level data fusion based on principal components (PCs) feature variables extraction is satisfactory with an accuracy of 100%. Hence, data fusion of FT-MIR and NIR signals can effectively identify the geographical origin of wild P. yunnanensis.

Discrimination of Gentiana and Its Related Species Using IR Spectroscopy Combined with Feature Selection and Stacked Generalization

“…In the section, all of the classification models were established by full spectra data (the total number of points in NIR and FT-MIR is 1487 and 1214, respectively) and 180 samples were separated into a calibration set (108 samples) and a validation set (72 samples) by the Kennard-Stone algorithm [48]. Six performance parameters, including sensitivity (SE), specificity (SP), efficiency (EFF), accuracy (ACC), Matthews correlation coefficient (MCC), and Cohen's kappa coefficient (K), were applied to evaluate the identification ability of classification models [49,50]. Those parameters values range from 0 to 1, indicating a perfect classification when the values are 1 [49].…”

“…Six performance parameters, including sensitivity (SE), specificity (SP), efficiency (EFF), accuracy (ACC), Matthews correlation coefficient (MCC), and Cohen's kappa coefficient (K), were applied to evaluate the identification ability of classification models [49,50]. Those parameters values range from 0 to 1, indicating a perfect classification when the values are 1 [49].…”

“…For RF models, model performance depends on the proper selection of the hyperparameters, which are n tree and m try [49]. Figures S1 and S2 show the suitable hyperparameters and variation of model mean misclassification error (MMCE) with different hyperparameters.…”

“…In the study, five methods were used to feature selection ( Figure 6). Firstly, VIP (variable importance in projection), Boruta, GARF (genetic algorithm combined with random forest), and GASVM (genetic algorithm combined with support vector machine) were applied to select feature variables [49,50]. Secondly, the intersection of feature variables that were selected by these four algorithms was calculated and the result was the fifth approach of feature selection (Venn selection).…”

“…Presently, the application of stacked generalization for establishing classification models of different medicinal plants or herbs is rather scarce. On the contrary, another modeling approach, data fusion strategy, has been widely used for classification and geographical origin traceability of herbs and foods [48,49,53,54]. Some researches stated that spectra data fusion, such as low-level and mid-level fusion strategies, could improve the discrimination capacity of the classification models and those strategies were usually more efficient than single spectroscopic techniques for modeling [48,49].…”