企业资本结构及其影响因素的关系研究--多元线性回归模型与神经网络模型的比较与应用

将神经网络模型引入企业资本结构及其影响因素的关系研究领域,并使之与传统的多元线性回归模型相比较,选取1051家中国上市公司2002年的横截面数据,以资产负债率为资本结构的表征变量．以规模、盈利性、成长性、非债务税盾、资产结构、资产流动性、产品独特性、产生内部资源能力等公司特征为自变量,对上市公司资本结构及其影响因素进行实证研究,并对两模型实证结果进行检验与分析,得出神经网络模型的SSE较小、预测能力较强的结论。

A Study on the Relationship Between Capital Structure of Enterprise and its Determinants --Comparison and Application of Multivariate Linear Regression Model and Neural Network Model

This paper introduces Neural Network(NNW) Model into the research of the relationship between the capital (structure) of enterprise and its determinants, and compares it with the traditional Multivariate Linear Regression Model. (Based) on the cross-sectional data in 2002 of 1051 Chinese listed companies, the paper selects total debt ratio as proxies of (capital) structure, and selects the corporation characteristic variables - size, profitability, grow opportunities, non-debt (taxation) shields, asset structure, asset liquidity, product uniqueness and producing internal resource ability as the (independent) variables. Then it makes an empirical analysis on determinants of capital structure, then tests and analyzes the results of these two models. Finally, it draws a conclusion: the SSE(sum of squared errors) from the NNW Model is (smaller) than those from Multivariate Linear Regression Model, and the NNW Model possesses superior predictive ability.