A new tutorial offers developers pointers on how to implement recommendation engines using Google's cloud computing and machine learning technologies.
Google released a primer of sorts for enterprises on how to use its machine learning and cloud platform technologies to build an online recommendation engine for their Websites.
It uses the example of a house-renting Website to walk developers through the process of creating an engine capable of suggesting houses that the user might be interested in based on previous searches and behavior.
The goal is to give developers an idea of how to use open-source technologies and machine learning to implement a simple product recommendation engine on Google's cloud platform, Matthieu Mayran, a cloud solutions architect, wrote in a recent blog post
The sample system used in the tutorial consists of a front end for capturing and collecting user interaction data and a permanent storage system for the data. It includes a machine learning component based on Google's Cloud Dataproc for managing Hadoop and Spark data sets and another front-end storage system designed to be used in real time by the front end that generates recommendations.
In addition to walking developers through the process of choosing the right components, Google's tutorial
offers guidelines on the different considerations they need to keep in mind, such as timeliness concerns and filtering methods, when implementing a recommendation engine.
The tutorial predictably touts the suitability of Google's technologies, such as its redundant data center infrastructure App Engine, Cloud SQL and expertise in big data technologies, such as MapReduce and Dremel, for handling the compute-intensive workloads that power recommendation engines.
"We hope that this solution will give you the nuts and bolts you need to build an intelligent and ever-improving application that makes the most of the information that your users give you," Mayran wrote in his blog.
The primer represents Google's latest attempt to get developers to harness its machine learning technologies in innovative ways. Last November, Google moved
TensorFlow, the second-generation machine learning technology behind some of the company's services, such as Google Translate and Smart Reply, to the open-source community.
Company CEO Sundar Pichai at the time had described the move as an attempt to spur research around machine learning by making it available to engineers, academic researchers, developers and hobbyists.
In addition to the recommendations engine primer, Google also released another tutorial
designed to give developers an idea of how the company's Cloud Platform and TensorFlow can be harnessed to deliver what it described as fast, interactive data analysis and machine learning using big data sets.
For this tutorial, Google made available about six years' worth of financial time series data from eight different stock markets that developers can query and run analytics against using technologies like Google BigQuery and Datalab.
The tutorial is designed to give developers an idea of how to use its cloud technologies to obtain and merge data from different markets, perform data analysis on the merged data set, and use TensorFlow to build and train models for predicting financial markets.