Machine Learning in Finance: Using Neural Networks to predict stock returns

There have been multiple attempts to predict stock returns using machine learning, which have largely used historical time series data on share prices to make these predictions. Those attempts create networks which only work on one firm\u27s data, and cannot be applied generally. This study uses a neural network to predict stock returns based on financial and economic data. The method that is employed here predicts whether a given stock will beat the S&P 500 index over a future time period. This method has reached prediction accuracy of 64.5%. A method which makes consistently accurate predictions helps to identify additional factors that determine a firm’s value beyond what is generally accepted in the literature