目的 比較不同模式識別方法結(jié)合化學(xué)指紋圖譜對不同產(chǎn)地三葉青Tetrastigma hemsleyanum的鑒別效果，提出三葉青產(chǎn)地鑒別的新方法。方法 收集浙江、云南和貴州3個產(chǎn)區(qū)的72批三葉青樣品，采集HPLC指紋圖譜，標(biāo)注18個共有峰，比較主成分分析、正交偏最小二乘法-判別分析和隨機森林算法在處理不同產(chǎn)地樣品復(fù)雜數(shù)據(jù)中的差異效果。結(jié)果 72批三葉青樣品經(jīng)主成分分析只能分為2類，正交偏最小二乘法-判別分析的結(jié)果優(yōu)于主成分分析，隨機森林法能將3個產(chǎn)區(qū)的樣品完全分開。利用隨機森林算法結(jié)合指紋圖譜能有效地對不同產(chǎn)區(qū)的三葉青進行鑒別和區(qū)分。結(jié)論 本研究可作為三葉青產(chǎn)區(qū)和質(zhì)量控制的有效方法，為多指標(biāo)的復(fù)雜指紋圖譜鑒別不同產(chǎn)地藥材提供有效的參考。

Objective To compare the identification effects of different pattern recognition methods combined with chemical fingerprints on Tetrastigma hemsleyanum from different habitats, and propose a new identification method for the habitats of T. hemsleyanum. Methods A total of 72 batches of T. hemsleyanum samples were collected from Zhejiang, Yunnan and Guizhou. HPLC fingerprints were collected, 18 common peaks were marked, and the difference of principal component analysis (PCA), orthogonal partial least squares-discriminant analysis (OPLS-DA) and random forest (RF) in processing complex data of samples from different habitats was compared. Results A total of 72 batches of T. hemsleyanum samples can only be divided into two categories by PCA, the results of OPLS were better than PCA, and the RF can completely separate the samples from three habitats. The RF combined with fingerprint can effectively identify and distinguish T. hemsleyanum from different habitats. Conclusion This study can be used as an effective method for the quality control of T. hemsleyanum from different habitats and provide an effective reference for multi-index complex fingerprint identification from different habitats.

浙江省公益技術(shù)應(yīng)用研究項目（2017C32075）；浙江省基礎(chǔ)公益研究計劃項目（LGN18B020001）；寧波市領(lǐng)軍和拔尖人才培養(yǎng)工程擇優(yōu)資助科研項目（NBLJ20180101）