MachineLearning is an extremely fascinating process that involves several disciplines of computer science and allows machines to learn in a systematic and precise way a useful behavior for the purposes of the client, whether commercial or purely technical.
Machine Learning: how do you train a machine?
What is crucial to keep in mind, when talking about machine learning, is the knowledge that you have to pre-process the data that the machine will have to analyze.
Starting from the exclusion of irrelevant data (and this requires a bit of AB Testing) to then prepare those on which to base our statistical analysis, without altering their strength.
We proceed, then, to match the variables and understand which are the correlated ones, so that we can divide them into pairs and finally prepare a real prediction and all possible actions.
Goodcode's experience in Machine Learning
We'd love to explain everything we know about Machine Learning and what better way than to tell you what we've managed to do? Here are two examples of how we applied Machine Learning in two very different areas*.
**The examples we will give will not mention the names of the companies involved due to confidentiality issues.
Machine Learning in the Web: how to predict the future behavior of a user or a segment of users.
Every Digital Creator's dream is to know, before publishing a piece of content, whether or not users will appreciate it and whether their site will benefit in terms of traffic.
This is absolutely possible. The next step after the simple analysis of past data (or even current, thanks to the Real Time Views that Google Analytics offers) is precisely the prediction of future results.
Data Analysis always allows, in any field, to dispel a lot of myths about what a Creator thinks. These are the famous "irrelevant data" mentioned above: a practical example?
We discovered that the time and day of publication of a content is absolutely irrelevant in the prediction, because each user (or cluster of users) is exposed to a series of infinite variables that do not compromise the result: whoever is going to access a content, will do it anyway.
The process of Machine Learning in the Web environment has allowed us to mainly
predict the behavior of users on a given content (Reach, Click, Views, accesses and Engagement) and help the Creator to better optimize the content proposals based on an analysis of the Trends and preferences of its Target.
Machine Learning as a workforce, aka, that time Goodcode created a Craftsman Robot.
The poetry of an ancient material, the rock, opposed to the symbol of the future par excellence, the Robot.
The extraction and processing of rocks is a widespread activity in our territory and how can a robotic arm do such a precise and meticulous job?
We developed a hybrid system of Machine Learning and Artificial Intelligence by installing, on the industrial robotic arm, an optical camera and a set of flashes.
Thanks to the lights, the robot is able to perceive the three-dimensionality of the object on which to operate, but not only: the machine recognizes all the shades of color typical of the rocks also managing to label the quality. This, clearly, in an environment that is defined as "uncontrolled" (we are still talking about excavations/building sites) but without affecting the result.
The goal? To automate an ancient and meticulous work thanks to an autonomous and performing machine: an operation that improves the workflow of mass production.