A production function estimation to input-output relations with support vector regression

Abstract

Support vector machines are used for classification with the input vectors using a decision surface into a high dimensional feature space. In this paper, the mostly known Cobb-Douglas production function is examined the input-output relations in a textile industry. A support vector regression (SVR) model is established to estimate the 17 different cost and rate values as input data. An ABC analysis is applied to input factors that only 5 of 17 are more important. Then, SVR model is estimated to output with a sensitivity analysis. The results are shown with the estimation error exceeds the defined lower and upper limits in approximately 7 data out of 60 data. At the same time, it is observed that the number of support vectors has decreased to 3. Consequently, the effective solutions with reasonable solution times are presented in this study, thus, the machine learning, deep learning and metaheuristics methods with SVR might be applicable as further research for the different industrial problems together.