• Machine learning: what is it?

    In 1997 the Deep Blue computer defeated the chess genius Garry Kasparov: the IBM system had been trained watching dozens of games to win the world champion.

    This milestone marked pre and post era Machine Learning (ML). ML is a branch of Artificial Intelligence that instructs machines to think and learn from experience.

    Learning is not just about storing and collecting data, but about creating a model or path that starts with the information collected. The systems learn by practice but, above all, can train, thus optimizing their behavior through data processing.

    ML is the technology behind the voice recognition of virtual assistants, thanks to which we receive purchase suggestions on digital platforms such as Spotify or Netflix, intelligent responses from Gmail or even allows Cabify or Uber to reduce the duration of each trip.


    Machine Learning: 3 ML algorithms

    ML takes place through a series of algorithms that analyze large amounts of data and determine the best result for a specific situation.

    There are various types of algorithms, but we can catalogue them into 3 macro-categories:

    1. Supervised learning algorithm. The machine learns from examples: it is equipped with labelled training data, and it has to detect a model to formulate forecasts. The results are corrected by an operator, and the algorithm can then make the changes and reach a higher level of accuracy.

    An example? The Facebook algorithm. This one analyzes available public data to improve its service.

    1. Unsupervised learning algorithm. In this case, the algorithm uses unlabelled data but looks for similarities in the information. Clustering algorithms run through data and find natural clusters if they exist.

    An example? The way Airbnb sorts accommodations by location.

    1. Reinforcement Learning. This is a mix of the two previous types. This system teaches the machine through trial and error. It learns from past experiences and adapts its behaviour to the situation in order to obtain the best possible result.

    An example? Deep Blue is learning to play chess. Not only does it recognize the correct moves for each chess piece, but it is also able to understand which one is best suited to follow a particular strategy.


    In which areas can we use ML?

    ML is useful to:

    – predict technological equipment failures

    – know the right time to publish posts on social networks

    – predict diseases according to each patient’s symptoms

    – provide customer support with intelligent conversations conducted by chatbots

    – determine the potential customer profile based on their browsing behavior

    – detect traffic in a city in advance

    – know the right time to send a newsletter or schedule a call

    – intercept crimes in a telecommunication network

    – forecast incidents within automated systems and robots

    Ask for a live demo

    Start using it now, and discover how to improve your revenues.