Foreign Exchange (FOREX) Price Prediction


First and foremost, please note that I am an absolute newbie to Tensorflow and machine learning. I am also a novice FOREX trader with amateur experience programming in python, MQL5, Java, Javascript, C++, and Dart.

I am interested in predicting the direction of price for foreign exchange, which is a time series problem in my understanding. I have had several ideas about how I could do this, however, please help point me in the right direction in terms of codelabs and documentation to study.

The latest idea I have had is to try and predict if the next candlestick (please google Japanese price candlesticks if you are unfamiliar with the concept) close price on a particular trading timeframe will be higher or lower than the current close price depending on the details (open, high, low, and close prices) of the previous n number of price candlesticks.

I thought that I could approach this problem using reinforcement learning by exposing an agent to time series data and attempt to predict whether the close price will be higher or lower than the current close price in the next timestamp of the series as I have already mentioned above, and then repeat the same process again for the next candlestick and again until the agent is good enough. This solution could also be great for binary options trading.

This in essence, according to my understanding, will be looking for price action patterns in the selected n historical price candlesticks of the selected trading timeframe to predict the future.

Please help me create an environment to expose an agent to historical price data, and an agent to train on this data. Please note that price data is structured exactly the same way stock price data is structured.

I think that you can start with an overview on:

A Survey of Forex and Stock Price Prediction Using Deep Learning

1 Like

Hello @Bhack,

Thank you. I will have a look.