The Prediction API enables quick and easy automation of repetitive tasks. It accesses Google’s machine learning algorithms to analyze your historic data and predict likely future outcomes. All you need to do is upload your data to Google Storage for Developers, then use the Prediction API to make real-time decisions in your applications.
Prediction API is RESTful in nature, and is fast for most complex decisions.
Google is now announcing that they are adding new sets of features which will make the Prediction API even more powerful.
Multi-category prediction: Previously, using the Prediction API, each article could only be tagged with one label – the most pertinent one. Now articles can be tagged with all labels, ranked by pertinence.
Continuous Output: A recommendation API, based upon relevant data.
It could be very tricky to actually recommend someone with a most relevant tip, based upon their historic data that has absurd randomness. Using the new continuous output option, your app can provide a fine-grained ranking of Tips based on how well they fit the user’s preferences.
Mixed Inputs: Automatic Decision-Making based on multiple inputs
Problems like making automatic moderator for your blog, is one of the best examples Google quotes. Classifying incoming posts automatically based on comment text, username of the poster, and number of times they’ve posted before or the number of users that have liked their posts (numeric inputs). Both numeric and text data can be incorporated in your moderation helper, greatly improving accuracy and letting you get back to making content rather than managing it.
Combining Continuous Output with Mixed Inputs: Both when combined can do some awesome stuff like : set thresholds for automatic posting, automatic rejection and manual moderation.