Any Use of the Google Prediction API for financial time series prediction models?

Guys can we use it for Forex?

I had this mail today from Google:

Hello There!

The Google Prediction API originally launched as a private beta at Google I/O 2010. Since then we had a fantastic launch at Google I/O 2011 as a Google Labs product and have released a host of new features: utility functions, confusion matricies, PMML support, data anomaly detection and more. We reached another milestone last week: Google Prediction API graduated from Labs with version 1.4.

We have made incredible progress in the last few months. We think Google Prediction API is now ready for prime time. So take it for a spin if you haven’t had the chance yet - sign up at http://code.google.com/apis/predict

If you are wondering what Google Prediction API is all about, watch this fun introductory video:

http://www.youtube.com/watch?v=u39rCNFWDEA

Happy Predicting!

-Zach Goldberg

on Behalf of the Google Prediction API Team

p.s. Some other helpful Google Prediction API Links:

Hello World Getting Started Guide (http://code.google.com/apis/predict/docs/hello_world.html)

Sentiment Analysis Application Tutorial (http://code.google.com/apis/predict/docs/sentiment_analysis.html)

Modelling Best Practices (http://code.google.com/apis/predict/docs/developer-guide.html#designrecommendations)

Hmmm................................................

Anyway I am not certain that is good idea. Because:

http://code.google.com/apis/predict/docs/general_discussion_forum.html

The question was:


"Hello,

I've just came by Prediction API which is something of a great interest to me. I run a company which focus on this kind of solutions, supplying clients with prediction/recomendation systems. Thus, I'd love to ask a few questions.

Actually the quality of training/prediction is still not amazing. Training process doesn't accept any parameters other than learning data itself so I am wondering what algorithms do you use (SVMs, neural networks, Bayesian networks)? You must be using some implicit learning parameters - how about exposing them to public in order to make learning more efficient?

I will keep an eye on this interesting project. Hopefully I might be of some help."


The answer was:
"The Google Prediction API auto-tunes its parameters to give the best results for your specific model, so you have less up-front work to do. Regarding the algorithms, we are constantly evaluating and improving their type, extent and performance so we don't comment publicly on the specific algorithms used by the prediction API.
Marc
Google Prediction API Team