It seems that the gadgets and equipment that we use today are getting smarter by the day. They are not only taking doing routine tasks but even helping out make decisions based on trends and recommended practices. The Google Prediction API allows you to tap into Google's Machine Learning algorithms that crunch data and give your possible outcomes, thereby helping you make your applications smarter. And car-maker Ford may even be using it.
At Google I/O, Ford Motors provided an interesting demonstration of how it has been using the Prediction API in its research lab to make driving more efficient. Ford tracks the driver and Builds a history of the places that the driver regularly goes to and the path taken. Based on this, the system is able to detect intelligently that if you are driving the car at a particular time on the weekday, it might most likely be a trip to the office. In case of its hybrid car, the system will be able to recommend and use more efficient driving mechanisms where it could shift to the electric mode in certain residential areas or areas where fuel emission restrictions are in place. It could also switch over to gasoline to save on the electric charge for certain segments of the journey. While this is still in prototype stage, it is easy to see how Predictive Analysis might make into mainstream a few years down the road. PCWorld has an in-depth look at Ford's use of the Google Prediction API
The Google Prediction API can be used for various other scenarios where you need to intelligently make decisions based on data that you have been gathering all along. For example, you can use it to build recommendation systems, spam detection, sentiment analysis, language identification and much more.
To get started with the Prediction API, you need a Google account and must enable the Google Prediction API and the Google Storage API in the APIs Console Project. Once you have done that, it is recommended that you read the Hello World application, where the Prediction API is used in a simple use case of determining the language that the email was sent in.
You need to upload and train the system first before you can engage with the model by sending it queries. The developer documentation has more details on the process. The Prediction API is REST based and various client libraries are available to interact with it. There is a free usage quota beyond with you are charged as per the pricing terms.
Several prediction models for language identification, tag categorization and sentiment prediction are already built and hosted, so you can try them out. Over time, we should see several models hosted here to take advantage of without going through the process of building data and making the system learn first.
The Google Prediction API allows developers today to tap into the vast amounts of data that is being generated and make their applications smarter through prediction. Ford Motors case of applying it to improve the usage of its hybrid vehicle is an example of what is possible. We should expect more use cases like this to emerge and the emergence of already trained and hosted predictive models is likely to boost developer participation.
Image Courtesy: CNET Car Tech blog.