Can value-based stock selection criteria yield superior risk-adjusted returns: an application of neural networks

This study examines whether superior investment returns can be earned by using neural network modeling procedures to perform forecasts based on a set of financial ratios reflecting traditional value based investment strategies. The study covers a 20-year period. We find that the value ratio provides useful information that permits the selection of portfolios that provide investment returns superior to the DJIA and the S&P 500, and a group of randomly selected securities. The risk-adjusted returns for the portfolios selected by the neural network are greater than those achieved using other forecasting methods.