ABSTRACT
Data envelopment analysis (DEA) has been widely adopted and applied in numerous research domains and reaches superior performance outcomes for-profit organizations as well as for non-profit organizations. Unfortunately, when it comes to handling large amounts of variables with respect to observations, its discriminant ability will be extremely degraded. That is, it cannot tell the differences between efficient and inefficient decision-making units (DMUs). If the performance evaluation model loses its basic evaluation function, it is impossible for decision makers to deploy resources to appropriate places as well as to give quick responses to market dynamics. To combat this, one of the dimensionality reduction techniques, called Gaussian process latent variable model (GPLVM) is considered. It can learn a low-dimensional representation of high-dimensional data via the Gaussian process and represent the inherent messages with less complexities. By joint utilization of GPLVM and DEA, the decision makers can point out some structures hidden behind successful practices. To demonstrate the effectiveness of the proposed decision framework, the pharmaceutical industry is taken. It is because this specific industry recently received a bunch of resources from public and private sectors and receives lots of market participants’ attentions, especially in today's pandemic outbreak era. The results indicate that the proposed model reaches superior discriminant ability.
