PREDICTING BANKRUPTCY IN ROMANIA USING ARTIFICIAL NEURAL NETWORK

Abstract

In this paper we will present the results achieved from our experiments which predict the bankruptcy of limited liability companies in Romania, using artificial neural networks. All information and data used were received from the Romanian Ministry of Public Finance and National Trade Register, the data being reported by companies in 2018 and 2019. The data set is mixed, consisting of both public and private data. The private data could be used following an agreement with the two institution mentioned above. The sample consists of both healthy companies and bankrupt companies, each company comprising a total of 17 variables to analyze. The result obtained was good, more precisely following the experiments performed, it resulted in an accuracy of 97.67% for training, respectively 96.27% for testing.