Machine Learning: this unknown being

Have you ever opened Netflix only to realize that the home page content precisely reflects your tastes?

Or have you ever seen that the shuffle on Spotify offers you those exact songs that you love the most?

By chance, you say? I really don’t think so.

But don’t worry, it’s nothing serious. You are just the victim of one of the many applications of a machine learning algorithm.

Today, I am here to explain to you what it is and, why not, try and make you become friends.

Are you ready? Let’s start!

What is Machine Learning?

An article by McKinsey describes it as: “a data analysis method that automates the construction of analytical models”. It is a branch of Artificial Intelligence that is based on the idea that systems can learn from data, identify models independently and make decisions with human intervention reduced to a minimum.

But let’s go back to its origins.

The term “Machine Learning” was coined by Arthur Lee Samuel in 1959 and comes from the theory that computers can learn to perform specific jobs without being programmed to do so, thanks to the recognition of patterns in data.

Through algorithms, which learn from the data, computers can identify information, even unknown information, without being told where to look for it.

From this, perhaps the most important aspect of machine learning can be deduced: repetition.

Indeed, the more the machines are exposed to the data, the more they will adapt to the problem to be solved in a completely independent manner. In this way, human behavior is replicated: the computer learns from previous processing (just like a person with experience) and produces results by making reliable and repeatable decisions.

In today’s infosphere, a term coined by Luciano Floridi to define the exponential increase of data at our disposal, having machine learning algorithms in our toolbox is becoming a necessary requirement to obtain the right information at the right time.

How does Machine Learning Work?

Now that you know what it’s about, let’s go and see which methods these machines work with.

Machine learning, according to Arthur Samuel still, works via two approaches:

  • Supervised learning: the computer is given examples to use as an indication to carry out the job required. In his article “Machine learning is fun”, Adam Geitgey describes it like this: “In supervised learning, you are letting the computer work out that relationship for you. And once you know what math was required to solve this specific set of problems, you could answer to any other problem of the same type!”;
  • Unsupervised learning: the computer is not given any preliminary information. Data are provided without any indication of the desired result. This second learning method focuses on finding patterns and models hidden within the information. It wants to identify a logical structure in the inputs without these having previously been labeled.

These are the main but not the only models of machine learning. You can come across at least another two:

  • Reinforcement learning: with this method, the computer is stimulated to act in a dynamic environment and to have to learn from its mistakes (identified by a punishment) to reach the envisaged objective (reward). System learning is guaranteed by a continuous routine that revolves around punishment and reward;
  • Semi-supervised learning: the data the computer is provided with are incomplete. Some will be equipped with interpretative patterns and others will not. This can be considered a hybrid model.



Where is Machine Learning applied?

As I said at the beginning, the majority of the time, you will have come into contact with machine learning mechanisms without even realizing it. Let’s take a look at some practical applications.

One of the first environments where you will have found it is within recommendation systems. These take advantage of machine learning, learning from the behavior and preferences of the users who browse websites or mobile applications. AmazonNetflix and Spotify are three excellent examples of this application. Have you understood now why you always like the TV series that Netflix recommends so much? Or why Amazon always knows what you are looking for?

Even email anti-spam filters, which continuously learn to intercept suspicious messages and to react as a result, are regulated by these learning models.

The same goes for the prevention of fraud, and data and identity theft. By learning about people’s habits and events, computers are able to identify anomalous behavior in real time and block it. I bet you like this technology more already, right?

Even search engine results are nothing more than the effect of non-supervised machine learning algorithms. When you type “how to...” (input) into Google, the algorithms provide a series of information (output) considered relevant to the search performed on the basis of the analysis of patterns, models and data structures. Vocal recognition or handwriting recognition techniques are other machine learning applications.

Last but not least: self-driving cars. These recognize the surrounding environment (with data collected by sensors and GPS) and align their actions to it through machine learning algorithms.

3 Machine Learning techniques you should know


Now that you have understood what these Artificial Intelligence algorithms are capable of doing, the moment has arrived to go and see how they do it”.

The first technique that I want to show you is regression. Regression algorithms are mainly used to make predictions on numbers. The value of the customer or their life cycle can be analyzed. Or, optimal price can be predicted for a product to maximize its profit.

Through the analysis and study of quantitative variables, light can be cast on the enormous quantity of data at our disposal to make more timely decisions.

A second method is classification. This allows you to find the data that belongs to a known class. In this way, you can understand, for example, which are the best customers for a company. It allows you to carry out what is known in marketing as segmentation with much more efficiency, or to find out how many people no longer come into contact with the services of a company and why they don’t.

The last machine learning technique that I’m going to talk to you about today is clustering. This methodology allows the computer to group similar elements into the same group, creating segments with common characteristics.

This is extremely useful when you have to logically represent a large mass of data to then be able to study the connections and differences.

I will take the opportunity to point out that Content Intelligence too, or rather, Artificial Intelligence applied to content, uses machine learning technologies to automatically classify corporate content: in this way, it makes content “intelligent” and, therefore, capable of becoming a receptor for the interests of users who view it, obtaining valuable data with which to “adjust the aim” of the company’s communication.

That’s all for today! I hope to have cleared up a few things about machine learning and to have enabled you to understand how often it is present in our daily lives.

Hasta la vista, baby!

(You got that I was quoting Terminator, right?)