The Google Prediction API
Team has been hard at work on Release 1.5, which is available now, with the following new
features:
Model enumeration. We’ve added the ability to list all
of your models via the trainedmodels.list
request. You can obtain the entire list in one response or you can iterate through a large
listing in pieces using the maxResults and
pageToken options.
Model analysis. We’ve added the ability to obtain more detailed
information about data and models via the trainedmodels.analyze
request, which returns information about the trained model’s output values, features,
confusion matrix, and other information.
Simplified get method. We’ve simplified
the output returned by the trainedmodels.get
request. Model analysis data that previously was returned by a get
request (e.g. the confusion matrix), is now returned by the new
analyze request, along with additional analysis data. The
get response now returns a simpler model description along with new
timestamps indicating when the model was inserted and when model training completed, which
should make it easier to keep track of model lifecycle.
New Google App Engine samples. We’ve created two new sample apps
illustrating how to use the Prediction API from App Engine, coded in Python and Java. These
samples show how to create and manage shared server OAuth 2.0 credentials, and how to make
predictions on behalf of any site visitors using the shared server credentials. The sample
code is available here and
a live version of the sample app is available here: http://try-prediction.appspot.com.
You can read more about the API details here.
The new release is available now via the HTTP RESTful interface and our various language-specific client
libraries. You can also experiment with the new Prediction API 1.5 interactively via
the Google
APIs Explorer.
We’re always looking for ways to improve the Prediction API so, as always, please let us
know about any problems or feature suggestions you might have. Happy
Predicting!
Marc
Cohen is a member of Google’s Developer Relations Team in Seattle. When not teaching
Python programming and listening to indie rock music, he enjoys using the Google Prediction
API to peer into the future.
