3 results about "Partial least squares analysis" patented technology

The invention relates to the field of liquor identification methods, in particular to a method for constructing a model for identifying authenticity of Maotai-flavor liquor, which is characterized by comprising the following steps of: 1, collecting carbon isotope data of a sample: measuring delta 13C values of ethanol and flavor skeleton components in the liquor sample, and generating a stable carbon isotope data set of the Maotai-flavor liquor; 2, analyzing and screening characteristic variables based on a partial least square method; 3, constructing a machine learning model: importing the screened feature variables into the machine learning model, and training the feature variables through the machine learning model; and 4, model application: based on the trained machine learning model, importing the characteristic variables of the to-be-detected wine sample into the model for identification, and obtaining an identification result of the to-be-detected wine sample. According to the method, the delta 13C values of the ethanol and the flavor skeleton components of the ethanol in the Maotai wine are measured based on the stable isotope mass spectrometry technology, the model is constructed, irrelevant noise is eliminated in combination with PLS-DA analysis, and the robustness, stability and accuracy of the whole model are improved.

The invention discloses a peucedanum praeruptorum dunn medicinal material quality analysis method based on a near-infrared technology, and belongs to the field of medicinal material quality analysis methods. The method comprises the following steps: collecting near infrared spectrum data of a radix peucedani medicinal material sample; preprocessing the near infrared spectrum data, and modeling by adopting one algorithm of a random forest, a back propagation neural network and partial least square analysis to respectively obtain quantitative analysis models capable of measuring at least one effective component in the content of praeruptorin A, the content of praeruptorin B and the content of praeruptorin C; obtaining the content of effective components in the radix peucedani medicinal material sample based on the model, and analyzing the quality of the radix peucedani medicinal material. According to the present invention, the near infrared technology is combined with the chemometrics to achieve the rapid determination of the three active components such as the praeruptorin A, the praeruptorin B and the praeruptorin C in the peucedanum praeruptorum dunn medicinal material; the sample preparation process is convenient, the sample is lossless, no chemical reagent pollution risk exists, and the result is accurate.

The application discloses a method for identifying the genuineness of radix codonopsis based on an electronic tongue and GC-MS technology, and discloses the following technical scheme: the electronic tongue technology is used to evaluate the tastes of radix codonopsis (Lu Dang, Bai Tiao Dang, Tai Dang, Wen Dang, Dao Dang and Tiao Dang) solutions of different commodities, principal component analysis and partial least square analysis methods are used to analyze the electronic tongue data of the radix codonopsis of different commodities, and it is found that the radix codonopsis of different commodities can be well distinguished through the taste identification of the electronic tongue. Through the primary metabolomics difference research of Lu Dang and other commodity radix codonopsis by GC-MS, 35 metabolites are identified and recognized, and PCA analysis and orthogonal partial least square analysis are conducted on the results, and 17 differential metabolites in the radix codonopsis are found. The Lu Dang and Bai Tiao Dang and Tai Dang of the same origin are analyzed, and 5 differential metabolites are found. The application can provide an efficient and objective analysis method for identifying the genuineness of radix codonopsis medicinal materials.