The recent revolution in app development is all about “predictive application” and it is estimated that more than 50% of apps on a typical smartphone have predictive features. For savvy companies , looking to metrics to analyse what happened in the past is not the only key to success. In addition, they develop models to predict optimal offerings in the future.
A cloud-based pattern matching and machine learning tool, the Google Prediction API, helps us to understand the requirements and habits of our customers. It is powerful and yet simple enough to not only track events on the app and collect usage data, but also to develop predictions and integrate them with the app.
The Google prediction API enables a new generation of smarter Apps and provides businesses with predictive analytics ability. These features can be applied in a wide variety of domains. Some examples include:
Customer sentiment analysis : mines the internet, mostly social media, to determine customers’ opinions and see if the comments have a positive or negative tone about a particular product or brand.
Churn analysis: analyses the loyalty of customers and the probability they may leave a brand or product.
Upsell opportunity analysis: based on customer's previous habits, this is a method to promote different upgrades or products.
e-mail classification: this predicts if an email is useful to read or not.
Smart Tagging: tags meta data using machine learning.
Product Recommendations: analyses which products your customers are likely to buy.
Priority Filtering: given a user's past viewing habits, predicts customer's priorities.
Spam Filtering: predicts if a comment or email is spam or not.
By reading the following examples you can see how the Google prediction API can be implemented:
This explains how a model can be built for predicting transaction revenue for each item.
This is a model developed to predict if a comment is positive or negative.
This app predicts your health status based on 1 of 6 input activities (walking, sitting, walking up stairs, lying down, etc.) .
This is an app that uses the Google Prediction API to predict fraud.
Ford takes advantage of Google's Prediction API by gathering data from a vehicle driver to take into account battery usage and optimise their power consumption.
In order to start with the Google Prediction API and Google's machine learning algorithms, firstly, you need a Google account and must enable the Google Prediction API and the Google Storage API in the API's Console Project. After that, it is highly recommended to read two simple use cases about purchase prediction and email classification in the Hello World application.
Then, go to developer documentation to learn how to upload and train the system. Once you have done this, you can look at samples and libraries. Some prediction models for language identification, sentiment prediction and so on are available in the prediction gallery.