浅谈数据包络算法在创新绩效评估中的应用与优化

Application and optimization of data envelopment analysis in innovation performance evaluation

数据包络算法(DEA)作为一种非参数前沿法,因其无须事先假设生产函数的优势,在多投入、多产出的创新效率评价中应用广泛,本文探讨了其在创新绩效评估中的应用。首先,文章详细介绍CCR、BCC、SE-DEA和SBM-DEA等几种常见的DEA模型的特点和适用场景。其次,以16个创新主体为对象,构建投入产出指标体系,包括3个投入指标、2个期望产出指标和1个非期望产出指标;分别使用CCR模型、BCC模型、SE-DEA模型和SBM-DEA模型进行评估并分析各模型的优缺点和适用场景。最后,文章指出DEA在创新绩效评估中的局限性,例如指标体系构建的不足和指标个数与DMU个数的关系等。针对这些问题,提出改进方法,例如对非线性指标进行预处理、构建多阶段DEA模型和优化指标体系等。

This paper explores the application and optimization of Data Envelopment Analysis(DEA)in innovation performance evaluation.As a non-parametric frontier method,DEA has been widely used in innovation efficiency evaluation with multiple inputs and outputs due to its advantage of not requiring prior assumptions about production functions.This paper introduces several common DEA models,including CCR,BCC,SE-DEA,and SBM-DEA,in the context of innovation performance evaluation,discussing their application scenarios and characteristics.Through a case study involving 16 innovation entities,an input-output indicator system was constructed,including three input indicators,two expected output indicators,and one unexpected output indicator.The case was evaluated using CCR,BCC,SE-DEA,and SBM-DEA models,and the results were analyzed,highlighting the advantages and disadvantages of each model and their respective application scenarios.Finally,the paper points out the limitations of DEA in innovation performance evaluation,such as the insufficiency of the indicator system and the relationship between the number of indicators and DMUs.In response to these issues,the paper proposes improvement methods,including preprocessing non-linear indicators,constructing multi-stage DEA models,and optimizing the indicator system.