Developing a Model to Predict Success of Agricultur al Production Enterprises Based on Their Capitals

Abstract

This study aimed to develop a recognition model in order to classify success of agricultural enterprises. To this end, the study in vestigated the relationship between capitals owned by the enterprise and the success le vel by using "Neural Network" model. This study was conducted during 2013-2014 in Zanjan County, Islamic Republic of Iran. Data was obtained through a structured questionnair e and holding interviews with 92 enterprise owners, out of 125, involved in producin g agrifood. According to the results of data analysis, Multilayer Perceptron Neural Network with Backpropagation algorithm was the appropriate algorithm to cope with the whol e scope of the study. Empirical analysis by SPSS indicated that the Multilayer Perc eptron consisting of one hidden layer with 6 nodes was an appropriate architecture. Class ification Accuracy Rate (CAR) and "Receiver Operating Characteristic (ROC)" curve wer e used to assess the model. Based on CAR of holdout data, the model was able to class ify 86.4% of the samples correctly. Also, the study intended to reveal the relative imp ortance of explanatory factors on enterprise success. Results indicated that human an d social capitals were the more influential factors, followed by economic and envir onmental capitals. Therefore, to promote agricultural enterprises, policy makers and managers need to improve software and hardware assets, simultaneously. Keywords: Classification, Neural Network, Perceptron, Pattern .