We’ve been applying machine learning to FOREX price prediction. The performance of our models varies widely, so to establish a baseline we created a simple linear regression model with which we can compare performance of more sophisticated models against.
What We Are Trying To Do
Given a time-series of 26 four-hour price samples, we want to predict the maximum price reached in the next ten four-hour periods. Schematically, after normalization, this looks like:
We create a linear regression model mapping the features to the max price reached. (The exact features we use are “secret sauce”—trade secret, for now). Running the model on a withheld test set produces the following known vs. predicted plot:
Not great performance; we hope to improve significantly upon these results using machine learning. However, cross-validation demonstrates that this model operates consistently, always a good thing: