研究目的
To establish a discriminant analysis model for classifying different kinds of medicinal liquor using FT-IR spectroscopy combined with SVM and PCA.
研究成果
The established model using FT-IR spectroscopy with SVM and PCA achieved high accuracy (validation accuracy 99%, prediction accuracy 97%), demonstrating its effectiveness for classifying medicinal liquors. This method is fast, simple, and accurate, with potential applications in quality control and authentication of traditional medicinal preparations.
研究不足
The study is limited to five specific types of medicinal liquors; other types or variations were not considered. The model's performance may be affected by impurities or noise in the spectra, and the generalizability to other beverages or conditions is not tested. Optimization of spectral processing and model parameters might be further refined.
1：Experimental Design and Method Selection:
The study used FT-IR spectroscopy combined with SVM and PCA algorithms for qualitative classification of medicinal liquors. The methodology involved spectral data processing, PCA for dimensionality reduction, and SVM for classification modeling.
2：Sample Selection and Data Sources:
Five kinds of Chinese herbal medicines (white ginseng, red ginseng, Chinese angelica, Chinese wolfberry, deer antler) were used to prepare medicinal liquors via impregnation method as per Chinese Pharmacopoeia (2015). A total of 576 infrared spectra samples were collected, with 476 for modeling and 100 for testing.
3：5). A total of 576 infrared spectra samples were collected, with 476 for modeling and 100 for testing.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: Equipment included Bruker Tensor 27 FT-IR spectrometer with ATR accessory, OPUS software, The Unscrambler 10.3 software, glass bottles, syringes, membrane filters (0.22 μm), anhydrous ethanol. Materials were the herbal medicines and base liquor.
4：3 software, glass bottles, syringes, membrane filters (22 μm), anhydrous ethanol. Materials were the herbal medicines and base liquor.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Liquors were prepared by soaking herbs in liquor at 25°C for three months. Spectra were collected in the range 4000–650 cm?1 with 32 scans, 10 kHz speed, 8 cm?1 resolution. Samples were filtered before injection into ATR, and the accessory was cleaned with ethanol between runs.
5：Data Analysis Methods:
Spectral data were processed using various methods (e.g., Smooth + Gap.dervatives + SNV), PCA was applied to select effective wave number bands, and SVM was used for modeling with grid search for optimal C and gamma parameters. Accuracy was evaluated based on training, validation, and prediction sets.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
