有向无环图在构建Logistic预测模型中的应用研究  

Research on the Application of Directed Acyclic Graph in the Construction of Logistic Prediction Model

作　　者：吴孟泽 Wu Mengze(Business School,Sichuan University,Chengdu,China)

机构地区：[1]四川大学商学院,四川成都

出　　处：《科学技术创新》2023年第3期63-66,共4页Scientific and Technological Innovation

摘　　要：传统的回归模型难以推断变量间的真实因果关系,可能导致研究结果偏离真实值。有向无环图(DAG)将变量间的因果路径可视化,根据后门原则可以定性分析变量间的因果关系。应用线性非高斯无环模型(LiNGAM)方法从观察数据中生成DAG分析变量间因果关系。从不同角度出发,提出两种基于DAG的Logistic预测模型变量选择方法,通过实验验证了新方法的有效性。Traditional regression models cannot infer the genuine causal relationship between variables,which may lead to the deviation of research results from the actual value. The directed acyclic graph(DAG)visualizes the causal path between variables and can qualitatively analyze the causal relationship between variables according to the backdoor principle. The linear non-Gaussian acyclic model(LiNGAM) generates DAG from observation data to explore the causal relationship between variables. From different perspectives,two methods of variable selection of the Logistic prediction model based on DAG are proposed, and experiments verify the new method’s effectiveness.

关 键 词：有向无环图 因果关系 Logistic预测模型 

分 类 号：O213[理学—概率论与数理统计] TP181[理学—数学]