stock-predict-by-RNN-LSTM

Karol Dzitkowski 's result is as follow.(High Frequency Trading Price Prediction

using LSTM Recursive Neural Networks, Karol Dzitkowski)

RNN avg err = 0 . 5 6 7 3 0 0 7 2 4 6 3 8 epsilon = 0.5, SGD optimizer

After I changed epsilon = 0.9375, optimizer to Adagrad optimizer

RNN ave err is as follow.

2nd RNN AVERAGE ERROR = 0.56443236715 RNN TEST ERRORS = [ 0.57669082 0.56219807 0.56944444 0.55193237 0.57789855 0.57246377 0.57125604 0.57729469 0.57246377 0.5839372 0.55012077 0.57125604 0.54770531 0.56038647 0.5531401 0.55736715 0.5736715 0.54770531 0.55495169 0.55676329]

Now, I need to incorporate Seq2Seq and Recurrent Shop in this program to tune up the performance at next step.