@Skate
1/ inputs Say I just use DJIA OHLCV end of day time serie data.
2/ formula There is no formula. The deep learning process is like an adaptative logic running inside a neural network. The process is like an avanced recursive learning from past to future.
Using formula is like constructing a model (ie forecasting the weather). You start from known data points and you propagate the scheme to other points thanks to mathématica formulas. This is NOT what I use and I think a Model is not the way to do because each day has its new rules. A mathematic Model cannot really adapt itself.
3/ the green line is the forecast, right. Each dot (the 3) represents the prediction of the close value, calculated from data one day before. so the last dot represents the next trading day close value forecast calculated from the last real known data.
4/ interpretation The green line must be interpreted as a day trend, not the daily range/magnitude which is rarely good. So the March 5th chart shows that the learning processed computed an up trend (green line from dot 1 to 2) for March 4th which is confirmed by the blue line (real data), and shows a down trend (green line from dot 2 to 3 and last) for the next trading day (March 5th).
But you have to be aware
a/ that the forecast trend goes is over the previous real value for an up trend,
b/ that the forecast trend goes is lower that the previous real value for an down trend.
Otherwise an interpretation is not possble.
1/ inputs Say I just use DJIA OHLCV end of day time serie data.
2/ formula There is no formula. The deep learning process is like an adaptative logic running inside a neural network. The process is like an avanced recursive learning from past to future.
Using formula is like constructing a model (ie forecasting the weather). You start from known data points and you propagate the scheme to other points thanks to mathématica formulas. This is NOT what I use and I think a Model is not the way to do because each day has its new rules. A mathematic Model cannot really adapt itself.
3/ the green line is the forecast, right. Each dot (the 3) represents the prediction of the close value, calculated from data one day before. so the last dot represents the next trading day close value forecast calculated from the last real known data.
4/ interpretation The green line must be interpreted as a day trend, not the daily range/magnitude which is rarely good. So the March 5th chart shows that the learning processed computed an up trend (green line from dot 1 to 2) for March 4th which is confirmed by the blue line (real data), and shows a down trend (green line from dot 2 to 3 and last) for the next trading day (March 5th).
But you have to be aware
a/ that the forecast trend goes is over the previous real value for an up trend,
b/ that the forecast trend goes is lower that the previous real value for an down trend.
Otherwise an interpretation is not possble.
Anyway you have to consider this system in conjonction with other kind of information
(Sorry for my english writting, my mother tongue is the french)
Be the first to comment