Data Science in Financial Investment

It has been proved that deep learning can present a great computational power, which can play an important role in financial investment. Applications of deep learning in investment are described below. Factors in the traditional factor model are normally linear factors, while non-linear factors can be extracted to help obtain better market information via deep learning. As linear factors tend not to capture all market information, non-linear factors can present a great role in this situation. Based on these characteristics, deep factor models are proposed recently. Furthermore, the deep factor model can be combined with a custom loss function to provide portfolios with better performances. Apart from this, deep learning can also facilitate obtaining some factors in factor models which has been proved empirically. For example, with the development of natural language processing techniques, text data from financial news and social media are analyzed to measure the market sentimental signals and predict the market trend. What’s more, discrete data from financial statements can also be used to build up novel models to allocate assets. Another important and attractive application of deep learning in the financial industry is financial data generation. That the data size is not large enough is a key problem when using deep learning to deal with some financial questions. However, based on the generative adversarial network, a one-dimension or higher dimensions time series dataset can be generated to increase the size of the financial dataset to avoid over-fitting. To summarize, deep learning can present a large influence on the researches of financial investment, which undoubtedly deserves us to pay much attention to exploring how to construct more efficient models with the strong computational power of deep learning.