Neural network based empirical models play an important role in virtually every aspect of the financial industry. The ability to learn which consumer income and purchasing habit attributes are significant when making decisions to grant credit has lowered the losses incurred by credit card companies both in issuing cards initially as well as in detecting fraudulent transactions after the card has been issued.

Similarly, the process of granting commercial loans or consumer mortgages has become highly dependent on empirical models that forecast the likelihood of repayment, again prior to issuing the loan, and then predict the likelihood of payment repayment problems or insolvency during the life of the loan.

In highly volatile securities markets, transactions in every type of financial instrument produce huge volumes of trading data that form the basis for neural network models to forecast everything from interest rates to the movement of equity indices or individual stocks. The many influences on institutional or individual investor behavior can never be quantified in a closed form algorithm, yet empirical models derived from the data consistently outperform human traders and portfolio managers.