融入Lasso的XGBoost组合优化方法及其中医药数据分析

COMBINATION OPTIMIZATION METHOD BASED ON LASSO AND XGBOOST AND ANALYSIS OF ITS TCM DATA

中医药物质基础实验数据不仅具有多成分、多靶点以及非线性的特点,还呈现出特征维数较高、样本较少的特性,而传统统计分析方法又难以解释非线性的数据,以及达到一个较好的降维效果。因此提出一种融入Lasso的XGBoost组合优化方法(LAXGB),该方法利用one-way ANOVA过滤无关特征,再结合Lasso算法进一步地去除无关特征和冗余特征,从而筛选出有效特征子集作为XGBoost算法的输入,再进行非线性回归,实现特征降维和模型优化的目的。采用中医药物质基础实验数据、中医药成分数据和UCI数据集进行分析,实验结果表明,该方法对中医药物质基础实验数据有较好的适应性。

The basic experimental data of traditional Chinese medicine not only has the characteristics of multi-component,multi-target and nonlinear,but also exhibits the characteristics of high dimensionality and few samples.However,traditional statistical analysis methods are difficult to explain nonlinear data and reach better dimensionality reduction.Therefore,this paper proposes a combination optimization method based on Lasso and XGBoost(LAXGB).One-way ANOVA was used to filter the irrelevant features,then the Lasso algorithm was used to remove redundant features,and the optimal feature subset was selected as the input of XGBoost algorithm.Then nonlinear regression was carried out to realize the purpose of feature dimensionality reduction and model optimization.The traditional Chinese medicine material basic experimental data,Chinese medicine composition data and UCI data set were used for analysis.The experimental results show that this method has good adaptability to the basic experimental data of traditional Chinese medicine.