Creditors carry the risk of their clients not meeting their debt obligations. In the literature, these events are often referred to as default events. These can be modeled for each company through a probability of default (PD). Measures can be taken to limit the default risk: in this paper we focused on credit limit. Firstly, we predict PD of a company using a logistic regression model, based on publicly available financial data. Secondly, we effectively find an optimal portfolio under risk aversion constraints and show how variation of inputs affects the results.