Is anyone using Gaussian Process models to build indicators. Something I'd like to investigate to model cycles.
Well, since nearly the entire options world is based on GBM, and GBM is based on Gaussian processes, I guess the answer for me and every option trader in existance is, yes.
the black model uses a normal distribution and can be derived using high school mathematics in an hour or so. Lets assume you have n orders placed and y are buy order and n-y are sell orders. If you clump enough buy orders or sell orders in a small enough time interval the stock will abruptly rise or fall by a certain percentage. Multivariate analysis can help you compute the probability of a resulting change in stock price using the via energy level equation. http://i39.tinypic.com/2vdo4r5.jpg
Thanks guys. I think GP models are great to do dimensionality reduction but I am not sure of their generalization capability in nonstationnary conditions. Any idea on that?