I know it's a tough issue to predict a stock prices but here are my project and problems:
First, I've trained the Network by providing 4 inputs (Open,Hight,Low,Close) and the expected output (Open,High,Low,Close) of the next period. I've ranged the prices by dividing them by 100 to be in the range (0-1). I didn't get fine results so far but I have some questions:
1- The error becomes static after some training trials. Why?
2- What's the best number of hidden layers?
I didn't find the right solution from the internet.