Year of Graduation
Predictability of Highly Liquid Assets Returns: Signs and Absolute Values
Predicting returns is a task that for many years remains relevant in the academic environment, since it is inextricably linked with the problem of determining the fair price of an asset. In this paper, we consider the use of machine learning methods (gradient and adaptive boostings, support vector method, etc.) in forecasting the weekly returns of the SP500 index and its futures. 110 technical and fundamental variables are used. The results of a stepwise forecast of returns show that during the last three years there is a significant opportunity to predict the behavior of the SP500 index itself and its futures (the correlation of the stepwise forecast of returns and actual returns is 31.8% for the index and 29.8% for futures), which exceeds Known results regarding predictability. This may indicate the existence of inertia of a nonlinear nature in price dynamics.