面向高不平衡高维混合数据的企业财务危机动态预警研究

Financial distress prediction (FDP) is a global problem and challenge. But it is an effective way to reduce the systemic financial risk. Other than traditional predicting data, the predicting data in the big data environment has complex characteristics for high-dimensional mixed, highly imbalanced, mass and dynamic data, in which the traditional FDP method based on statistical analysis and machine learning would not be able to obtain reliable results. This project plans to take the complex predicting data as the breakthrough point to develop a dynamic predicting method of financial distress based on the soft set theory for its great potential on analysis of big data..This project starts focusing on the influence factor analysis of financial distress in the big data environment to construct the initial FDP indicator set and extract the data characteristic..Secondly, sampling method, indicators selecting method, static integrated predicting model and dynamic incremental predicting model are established by expending soft set theory for the complex predicting data..Finally, we proposed the evaluation indicator and the application of predicting models to predict financial distress. Empirical research is employed to verify the performance and feedback the establishment of predicting models, to improve the ability of FDP under the era of big data..This project is committed to establish a dynamic predicting model of financial distress towards high-dimensional mixed data and highly imbalanced samples. It is not only enriches the research of soft set and big data analysis technology, but also provides a reliable theory and method for FDP in big data environment.

企业财务危机预警是世界性难题，也是防范、化解系统性金融风险的有效途径。不同于传统预警数据，大数据环境下的预警数据具有高维混合、高不平衡、海量及动态等复杂特征，传统预警方法无法进行有效预测。课题拟采用具有大数据分析优势的软集合作为理论基础，聚焦预警数据的高维混合特征，同时考虑其高不平衡、海量及动态特征，研究面向高不平衡高维混合数据的企业财务危机动态预警方法及其应用。首先研究大数据环境下企业财务危机影响因素，构建初始预警指标集，提炼数据特征；其次针对复杂预警数据，拓展软集合构建预警样本选取方法、预警指标选取方法、静态集成预警模型及动态增量预警模型；最后设计评价指标与应用方法，通过实证反馈改进预警方法，提高大数据环境下的预警适用能力。本项目面向高不平衡高维混合数据建立企业财务危机动态预警方法，不仅丰富了软集合与大数据分析技术研究，而且为大数据环境下的企业财务危机预警提供了可靠的理论与方法基础。

针对大数据环境下的企业财务危机预警数据具有高维混合、高不平衡、海量及动态等复杂特征，以具有大数据分析优势的软集合理论作为理论基础，开展面向高不平衡高维混合数据的企业财务危机动态预警方法及其应用的研究。根据项目计划的要求，针对大数据技术发展和应用对企业经营管理带来的变化，完成了大数据环境下的企业财务危机初始预警指标分析；针对大数据环境下企业财务危机预警数据高维混合、高不平衡、海量及动态等复杂特征，完成了面向高不平衡高维混合数据的企业财务危机动态预警方法构建；结合国内外不同规模企业的数据，完成了面向高不平衡高维混合数据的企业财务危机动态预警方法的应用及评价。从而较为系统地建立了一套面向高不平衡高维混合数据的企业财务危机动态预警理论、方法和技术体系。通过上述研究，目前已完成学术论文17篇，其中已经正式刊发9篇（2篇为SSCI检索期刊，4 篇为SSCI/SCI双检索期刊，1篇为EI检索期刊，2篇为CSSCI核心期刊），正式出版学术专著2本。基本完成了项目计划书规定的研究内容，部分实现了预期的研究目标和研究成果，并为后续开展相关深入研究奠定了较好的基础。