What is machine learning, how does it work and what is it used for

Expert post

Héctor Vicente Moya, Project Manager at PaynoPain

We have mentioned in more than one occasion that our payment gateway (Paylands) uses artificial intelligence like machine learning technology in its anti-fraud system so that it can "learn" and be increasingly effective. But what does this mean and how does this technology work? Learn about it on this week's expert post, a contribution more technical than usual for those who wish to better understand this fascinating and current concept.

What is machine learning?

Machine learning provides the technical basis of data mining. It is used to extract information from large amounts of data, expressing it in an understandable way so it can be made use of for a variety of purposes.

This means the machine learning process consists on finding and describing structural patterns in data. This way we can draw conclusions that would otherwise be impossible to obtain through regular data analysis.

Additionally, machine learning uses computational methods to "learn" from the data, giving increasingly accurate results. The rate of learning and its ability to do so at all is determined by the type of machine learning. It is even possible to process data without providing a model to follow.

Machine learning: how it works and what is it used for

• Supervised learning: finds patterns and develops predictive models using both input and output data. There are two types: classification and regression.
• Unsupervised learning: finds patterns based only on input data. This technique is useful when you aren’t sure what to look for. It is used mainly for the analysis of raw data.

• Virtual personal assistants (Cortana, Siri, Bixby, Google)
• Weather forecasts for next week
• "Win predictor" for sports (shows who has the best chance of winning)
• Medical diagnostics
• Predictive searches in Google
• Custom ads
• Security and anti-fraud systems (like the one we use in Paylands)

The results of some of these applications are more obvious to users than others. For example, surely most of us will have noticed predictive searches on Google or personalized ads, and have used a virtual personal assistant. Sometimes we can even feel that our applications know everything about us, because that is how automatic learning works, showing that, by collecting our daily data, it can predict any of our behavioural patterns and, in this way, ease our daily tasks.