面向纵横数据流概念漂移的企业财务困境预测动态建模研究

项目名称： 面向纵横数据流概念漂移的企业财务困境预测动态建模研究

项目编号： No.70801054

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 交通运输

项目作者： 孙洁

作者单位： 浙江师范大学

项目金额： 18万元

中文摘要： 基于企业财务困境预测的迫切现实需求，针对国内外理论研究的静态建模思想与企业运营动态性特点的严重脱节问题，本项目从财务信息纵横数据流和概念漂移的全新视角，对企业财务困境预测动态建模理论方法体系展开探索性研究和实证检验。研究内容：(1)财务困境预测动态建模的理论基础研究；(2)重点探索适合企业动态运营特点的财务困境预测动态建模方法体系，即构建融入样例选择与样例赋权的多分类器动态集成方法来处理财务困境预测动态建模的概念漂移问题，从建模速度和动态增量学习两个角度展开财务困境预测动态建模的支持向量机学习算法研究，提出在建模过程中动态选取财务困境预测变量的指标体系构建方法；(3)从单个企业纵向数据流和横向截面样本数据流两个方向展开财务困境预测动态建模实证研究。该研究将增强财务困境预测模型在时间推移过程中对财务困境概念漂移的适应能力，提高预测准确率，为企业财务困境管理机制及预警系统的建立提供重要工具。

中文关键词： 财务困境预测；概念漂移；动态建模；纵横数据流

英文摘要： Despite the practical urgency of business financial distress prediction, the theoretical research home and abroad is focused on static modelling, which disjoints with the dynamic operation of enterprises. This project explores the theory and method system of dynamic modelling for business financial distress prediction and empirically tests its validity. Its main research contents include: (1)basic theory of financial distress prediction based on dynamic modelling; (2)the method system of dynamic modelling for financial distress prediction that adapts to business dynamic operation, namely, dynamic ensemble method of multiple classifiers integrating sample selection and sample weighting to dispose financial distress concept drift, support vector machine algorithm for financial distress prediction that is improved from both aspects of modelling speed and incremental learning, and method of indicator system's dynamic construction in the process of financial distress prediction modelling; (3)empirical research on dynamic modeling method for financial distress prediction, which is based on both vertical data flow of single enterprise and honrizontal data flow of sectional samples. The research will promote the financial distress prediction model adatpting to the financial distress concept drift with the time moving, and improve the prediction accuracy. It also provides enterprises with important tool for building financial distress management mechanism and early warning system.

英文关键词： financial distress prediction; concept drift; dynamic modelling; vertical and horizontal data flow