To be announced

**Applications
of neural networks to finance**

"Classical" approach of Finance is using models based on partial differential equations (Black-Scholes type equations) and stochastic processes (Ito difussions) for more than 4 decades. It is worth to remark that none of these models is an ultimate correct model of the markets. Contrary to the laws of physics, which stay unchanged forever, the models that govern financial markets tend to get obsolite relatively fast, and hence, their need of continous improvement. This project advances the idea that an ultimate model of financial markets might involve a sophisticated NN trainned on the current available market data.

Furthermore,
the classical methods are not efficient when trying to *finding alpha*,
or the signal in the market, on which the trader should bid to make a profit.
It is our belief that the hidden signals, or irregularities in the market
can be unveiled by using deep neural networks, including both CNNs and RNNs.

The project involves using NN to study the following problems:

**Pricing options and bonds:**Closed form formulas for options (Black-Scholes-Merton type formula) are known since 1970s. However, they depend on some parameters, which sometimes are hard to estimate, and are valid only under certain assumptions (lognormality of returns, constant volatility, etc), which are not satisfied in today's markets. Trainning a NN on a stock values and corresponding option prices can lead to a NN that can price new options better than classical models.**Forecasting volatility:**There are sustained efforts in pricing derivatives under a stochastic volatility model. These include Heston, Garch, ARMA, etc models. Most of them do not provide closed form solutions and are relatively hard to apply in practice. This project explored the extend to which the use of an RNN can forecast volatility better than classical models.**Forecasting stock prices evolution:**Finding a signal in the stock market is desired by all market speculators. This project investigates the capacity of NN of forecasting stock prices.

Team Members:

Ovidiu Calin | Paul Brown | Aaron Bolton | Cedric Bernard |

Cristian Paul Bara | Oana Ignat | Hai Tran Bach | Dan Capps |