Using a Neural Network to discover optimal short-term opportunities in Forex-Trading
- 0 Collaborators
The goal is to design and train a LSTM Neural Network, which at a given moment of time can successfully predict the optimal trading strategy for a certain time horizon. I face the usual challenges that one faces when dealing with Neural Networks: choosing the right inputs, the right outputs and the correct network structure. ...learn more
Project status: Under Development
Overview / Usage
On eventual success, the results can be used for profitable trading.
Methodology / Approach
The trading opportunities discovery is based on time series forecasting with NN.
I concentrate on RNN, in particular LSTM.
I experiment with different inputs. I try to use inputs that are more general than the price itself. See if I can ommit the price data completely.
The main focus of my efforts though is to experiment with the outputs of the NN: not just price, but some attributes of a trading strategy. My hope ist that using gemeralized outputs can help the NN generalize better itself.
Technologies Used
TensorFlow with Python and C++