Superior Stock Selection based on Deep Learning ? - Evidence from the Cross-Section of Returns

The project aims to answer the question if advances in machine learning can be useful in predicting financial returns. Recent proceedings in deep learning show promising results in other research fields. The project contribution to the finance literature would be an in-depth analysis with respect to the cross-section of stock returns. More generally, it would contribute in how far deep learning is suitable for prediction problems, confronting very noisy data. Inspiration as well as crucial for the implementation of the project is the fact, that Google recently open-sourced its machine learning library TensorFlow.