Posted by Travis Green, Product Manager, Prediction API Team

Editors note: Today we’re at Google I/O, and we have a few announcements about how we’re making our developer tools more available and better for businesses. The third and final, below, summarizes important improvements to the Google Prediction API.

Last year we introduced the Google Prediction API to give select developers access to Google’s sophisticated machine learning algorithms to analyze a wide range of example data and provide predictions for likely outcomes. Whether you want to create an automated spam and security filter, provide recommendations based on customer demographics, or find hidden patterns in financial data, the Prediction API can quickly make your app smarter.

Today we’re making the Prediction API available to everyone. We’re also announcing some new features to make it more helpful to businesses including:
  • Streaming training data that allows for rapid adaptation to real-time data and allows refinement of predictive models.
  • Improved user interface that provides consistent prediction scores, error reporting, and simplified formatting.
  • 99.9% uptime service level agreement for additional assurance.
  • Gallery of pre-built third party prediction models (coming soon) that will enable the immediate deployment of established models without the need for original development, beginning with these demo models.
There are many use cases for the Prediction API, including several being developed by Ford Motor Co. Research. Ford is working to use the Prediction API to optimize plug-in hybrid electric (PHEV) vehicle energy efficiency and usage by predicting driver destinations and optimizing the driving controls based on driving conditions to conserve fuel. Using the cloud-hosted API, Ford has been able to access computationally-intensive machine learning algorithms from Google to find patterns that rank potential destinations based on previous driving paths.

The Google Prediction API is the only cloud service that allows businesses to access this depth of sophisticated machine learning with only a few API calls and by utilizing the flexibility of Google Storage. Learn more on the Google Code Blog or visit the Prediction API site to sign up for an account.