Using LSTM Neural Network for Time Series Predictions in Financial Markets

The article will test if Long Short Term Memory (LSTM) neural networks are suitable for high frequency foreign exchange (forex) trading. Major world currencies often correlate and affect each other. We will try to take advantage and we will feed time series data of several currency pairs and do correlation analysis on them. When traders do technical analysis they inspect different time periods in order to get better understanding on currency trends. We will also test if feeding the neural network with different time periods will lead to better predictions.