Businesses will soon be able to perform the same kinds of big data analysis that allows Amazon to recommend books, video games and other toasters to their customers. Amazon has announced the availability of Machine Learning, a fully managed cloud service designed to draw useful information from mountains of data that it is sometimes difficult to exploit, for reasons of complexity, time or energy.
Following in the footsteps of its closest competitors Microsoft, which announced Azure Machine Learning and IBM with Watson; Amazon Machine Learning technology allows any developer platform using historical data to build and deploy predictive models.
AWS seeking to do with the latest Machine Learning offering is to bring big data analysis to companies that need to know, for example the color of tennis shoes that sell best in New England, what process business is the most effective or what action on the social networks can generate the most loyal customers. To simplify the work of users with the data already stored in the Amazon cloud, Machine Learning is integrated with the Simple Storage Service (S3), Redshift and relational Database (RDS).
Amazon relies on big data analytics to operate its business model. There is the analytical works behind to predict what people want to buy tomorrow or to inform users of what others bought. The combination of machine learning and big data can bring their indicators and specific data they would never have considered before.
To be implemented correctly, the new instrument requires three preliminary steps. The first is the construction of the predictive model using the Amazon Simple Storage Service (S3) or with Redshift. Later, the skeleton created must be validated and optimized, and then be tested and compared with other models.
The service is designed for all developers who do not have expertise in the field of machine learning, providing them with a set of tools for developing intelligent applications focused on data analysis. Leveraging the new offer of Amazon Web Services, developers can use the AWS Management Console or the API to quickly create all the models they need, and generate predictions with high throughput, without bothering to buy hardware, deploy and scale the computational load, manage dependencies or monitoring and troubleshooting infrastructure. There are no installation costs and developers pay only what they use.
Amazon Machine Learning is based on the same scalable technology used for years in the community of data scientists of the company’s e-commerce, capable of generating billions of predictions per day and make them available in real time with high production volumes. Amazon Machine Learning facilitates obtaining predictions for applications using APIs without having to deploy custom code to generate the prediction, or manage any infrastructure.
As per Amazon, these models can be used for different objectives, such as detect fraud, thus preventing the churn rate of customers, improve customer service, content personalization, propensity modeling for marketing campaigns, document classification, and automated solution recommendation for customer support. Based on a highly scalable and already proven technology, machine learning is used by developers to generate more than 50 billion forecast a week.
Amazon’s latest move to machine learning is an interesting move because with it comes to compete with similar proposals made by Microsoft and Google. It can even affect the strategy of IBM, which works even larger scale in this analytical groundwork and has invested vast amounts of money on hiring mathematicians and statisticians to provide such services.