Data Envelopment Analysis and Multivariate Calculus to Assess, Classify and Predict the Productive Efficiency and Innovation of Companies in the Chemical sector

Abstract

A method to establish assessment, classification and prediction criteria to evaluate the productive efficiency and innovation of the companies in the chemical sector of Barranquilla was developed. Concepts of data envelopment analysis, discriminant analysis and artificial neural networks were used. Information associated with variables of labor climate, information management, knowledge management, productivity management and innovation, were collected. The results were then validated with the discriminant analysis and forecasting processes and prediction of the efficiency of the companies in the sector with artificial neural networks were modeled. The results show that the average efficiency in the sector is 52.9%, with 6 companies classified as efficient ones. With the multivariate discriminant analysis techniques, the classification quality could be determined, achieving a 92.6% correct classification. Likewise, the selected neural network model generated a classification accuracy of 98.82%, 95.78% and 94.28% for the training, test and reserve samples, which shows the relevance of the classification model. It is concluded that the analyzed variables are significant to discriminate productive efficiency and innovation.