Data

Create feature vectors as inputs for the NN, and a classification (potential profit) as the output.
Starting point for the training set. Use the format "YYYYMMDD HH:MM:SS" or enter a UNIX timestamp.
Number of seconds in each period. Each period is represented by a candle (open, high, low, close).
Size of the neighborhood around min and max.
Length (in number of periods) of the time period to collect patterns from.