Determining the creditworthiness of potential bank customers is a crucial step in the process of providing loans. An example of applications of neural networks in this article is executed on a specific group of construction companies operating in a specific time period in the South Region of Czech Republic. Article seeks a neural structure that will be able to classify each company based on the input of financial and non-financial data to a group of creditworthy companies, companies that will bankrupt in a year or companies that will bankrupt in the future. The result is the identification of the model with the greatest predictive power that can be used in practice. Its accuracy is around 96%.