A DEA integrated grey factor analysis approach for efficiency evaluation and ranking in uncertain systems

• DOI: 10.1016/j.cie.2021.107681
• Publication Date: 2021-10-07T21:39:13Z
• Authors: Mengdie Huang; Tangbin Xia; Zhen Chen; Ershun Pan; Lifeng Xi
• Keywords: Rank (graph theory); Adaptability; Rank correlation; Sample (material)

Abstract Efficiency evaluation and ranking in manufacturing systems involves complex and multiple input/output criteria. As a well-established non-statistical method for relative efficiency evaluation, data envelopment analysis (DEA) is questioned for the lack of discriminatory capability and the sensitivity to measurement error. Traditional multivariate statistical analysis used for multiple criteria performance evaluation has the prerequisite of underlying hypotheses and shows poor resolution facing up to insufficient amounts of sample data. We consider that the grey theory is a systematic analysis methodology that focuses on information insufficiency and model uncertainty. Thus, this study combines the advantages of DEA, factor analysis (FA), and grey theory to deal with the deficiencies. A DEA integrated grey factor analysis (GFA-DEA) efficiency evaluation and ranking model for uncertain systems is established. It is structured on the absolute degree of grey incidence (ADGI) matrix instead of the correlation matrix. Therefore, the underlying factors that explain a substantial proportion of the variance will be extracted, and the overall efficiency of all the identities will be evaluated and ranked. Furthermore, the model is employed to evaluate and rank the utilization efficiency of organizational quality infrastructure (QI) in 22 manufacturing enterprises in China, which is an important driver of business and operational performance improvement. According to Spearman rank correlation, the proposed method gives a higher correlation with super efficiency compared to other methods. The results indicate the adaptability of the GFA-DEA algorithm, which is able to offer a computationally easy and reliable result for small and uncertain data sets.