The Prediction of Earnings Based on Support Vector Machines

In this paper we attempt to apply a relatively new learning algorithm, support vector machines (SVM), to the future earnings prediction problem and expect to improve prediction accuracy by adopting this new algorithm. By determining what information was actually used by expert financial analysts, these studies can help users capture fundamental characteristics of different listed companies. The result shows that there is some significant improvement with the SVM over the logic model, and the problem can be perfect solved with the new algorithm.