Key Factors of Quality Formation of Fritillariae Cirrhosae Bulbus： An Analysis Based on Group-Lasso and XGBoost
10.13422/j.cnki.syfjx.20251614
- VernacularTitle:基于Group-Lasso与XGBoost分析川贝母品质形成的关键因子
- Author:
Panpan WANG
1
;
Zhenyu LIU
2
;
Yuhan WANG
2
;
Xueyang REN
2
;
Kai WANG
3
;
Guoxin LAN
1
;
Nong ZHOU
1
Author Information
1. College of Environmental and Chemical Engineering， Chongqing Sanxia University of Science and Technology， Chongqing 404120，China
2. Green Planting and Deep Processing of Genuine Medical Materials in Three Gorges Reservoir Area， Chongqing Engineering Laboratory， Chongqing 404120，China
3. College of Pharmacy， Chengdu University of Traditional Chinese Medicine， Chengdu 611130，China
- Publication Type:Journal Article
- Keywords:
Fritillariae Cirrhosae Bulbus;
total alkaloids;
Group-Lasso;
variable screening;
XGBoost
- From:
Chinese Journal of Experimental Traditional Medical Formulae
2026;32(23):219-227
- CountryChina
- Language:Chinese
-
Abstract:
ObjectiveTo improve the quality of artificial cultivation of the source plant of Fritillariae Cirrhosae Bulbus （FCB）， an XGBoost regression model was constructed based on the Group-Lasso method to screen the characteristic variables. The key factors affecting the quality formation of the source plant were then analyzed. MethodsFirstly， with the total alkaloid content （TAC） of FCB as the dependent variable， explanatory characteristic variables were screened from the influencing factors of TAC in the literature studying FCB quality from 2010 to 2024 based on the Group-Lasso regression algorithm. The screening results were verified by ten-fold cross-validation. Secondly， the present study established an XGBoost regression model to analyze the influencing factors of the quality formation of the source plant， and calculated the importance score of the explanatory characteristic variables. ResultsGroup-Lasso regression removed 2 numerical independent variables and retained the remaining 5 categorical variables and 6 numerical variables； cross-validation also showed that the number of screening variables is 11， with the corresponding optimal λ parameter value of 0.000 515 225 1. The results of XGBoost regression model confirmed the geographical origin as the most critical factor affecting the quality formation of source plants， followed by growth status and soil type， and then by mean temperature of the warmest month， original plant type， mean annual temperature， and altitude. Climate type， precipitation， frost-free period and relative humidity of the geographical origin exerted little effect. ConclusionThe prediction tools of Group-Lasso and XGBoost regression analysis models were preliminarily established for the quality formation of FCB. These models overcame the multicollinearity among numerical independent variables， and explored the geographical distribution of TAC and its key driving factors. The suitable cultivation areas can be selected or wild-tending bases established based on these key factors， thereby effectively avoiding blind introduction and cultivation. Finally， the present study provided a reference for the development and utilization of FCB， a theoretical basis for the market expansion and quality regulation， and a direction for the sustainable development and protection of new and efficient FCB resources.
