ForexBot

Contains code and data for machine learning-based forex trading strategy

Based on the initial analysis of the minute-by-minute and daily data for various currency pairs, a classifier based on the previous n price movements performs poorly, ie not significantly better than a random model. To further pursue classification based on previous n price movements would require better time resolution, ie ms data.

The next version will use spectral data - FFT and Shannon entropy - to extract features and classify.