I was trying to put my hand on neuronetwork with synaptic.js I created a Perceptron with 3 layers(5,15,1) and parsed historical data from 2007 to 2017 for 6 currencies.
- Australian
- Belorusian
- Euro
- Japan
- Swiss
- USD
I got currency historical data from Central Bank of Russia -> CBR
I used first 5 as inputs to the NET, and USD as a output for training. Supposedly the NET should predict current currancy rate USD/RUB As a result, a got:
Training started with 2481 sets for training
First training input: { input: [ 21, 35, 11, 21, 21 ],
output: [ 0.30120481927710846 ] }
Training ended
Current USD value supposed to be 59.24262096102017 RUB per 1 USD
At this moment current Rate RUB/USD is 59.171 :)