By Ricardo Galante, Customer Care and Analytics Specialist at SAS
The world is constantly evolving. But it is possible to say that many of the technological solutions used today are the advancement of technologies that our parents and grandparents used in the past. Analytical applications that, in general, were aimed at the academic segment, currently help us in everyday situations. For example, when applying for a financial loan, we are not aware that an analytical technology is used to authorize, or not, credit to an individual. Personal information is compared with that of thousands of other customers who have gone through the same situation. In general, if this individual's data and habits are similar to those of people who have honored their commitments, he will have the credit released.
But, how to identify this similarity between thousands of people and one individual in particular? How to avoid, as far as possible, the injustice of denying credit to customers who would honor their commitment to pay on time or releasing the requested amount to a person who would not honor it? The answer to these questions is found in the 'Analytical Models'. There are several strategies to create these models and one widely used is Machine Learning.
What is Machine Learning?
In the past, this type of methodology was something distant and more like science fiction, but today it is fundamental for business. Machine Learning is an area of Computer Science, created from research related to Artificial Intelligence. The increasing use of these methods is totally related to the strong computational growth. If in the past people tried to solve complex formulas almost manually, today technology is used to automate the construction of analytical models that use algorithms to learn from data in an interactive way. That is, the idea is that these algorithms are almost self-sufficient with as little human intervention as possible.
What makes Machine Learning different from Statistics or Statistical Models?
Machine Learning and Statistics are equivalent in some ways. This means that the results obtained with one methodology can be compared to another. However, when it comes to creating Statistical Models, the objective is to learn something related to the data, that is, to obtain insights from existing information. With Machine Learning, instead of simply trying to understand what happens in the data, what you have is the 'creation of examples' or 'rules', and with each execution of the algorithm, it is able to learn and improve from the examples. The interactive aspect of Machine Learning is important because models improve by learning from internal calculations producing better results.
Importance of Machine Learning Today
Machine Learning has historical importance. There are reports that this set of techniques was born during the Second World War and has been improved until the present day. To give you an idea, this methodology was already present, for example, in SAS since the 80's. At that time, through Machine Learning techniques, people were already looking to find behavior profiles in order to identify fraudsters.
The extraordinary impulse in the use of this methodology is due to a few reasons. First, because the amount of data is increasing at an unprecedented rate. For most organizations, the challenge is extracting valuable information from massive volumes of data from a wide variety of disparate sources. Another point is the storage options for this information – currently more accessible or even free. Also, computational processing power has never been cheaper or more powerful.
This means that with the right data, the right technologies, and the right analytics, you can quickly produce models that can analyze a large amount of data, regardless of its complexity, and deliver results in less time, more accurately, and with the minimal human intervention. The result? High-value forecasts that can guide better decisions.
Applications for Machine Learning are numerous, such as: fraud detection; online offer recommendations; real-time web and mobile advertisements; sentiment analysis based on textual sources from social networks; credit scoring; equipment failure prediction; new pricing models; network intrusion detection; writing pattern analysis; among others.
We can say that the central idea is to use a high performance technology, with an interactive aspect. What makes the use of Machine Learning advantageous is the possibility of creating systems that learn from their own data, obtaining the best results with minimal human intervention, generating highly effective results and enabling strategic decisions, even in real time.
But, how to identify this similarity between thousands of people and one individual in particular? How to avoid, as far as possible, the injustice of denying credit to customers who would honor their commitment to pay on time or releasing the requested amount to a person who would not honor it? The answer to these questions is found in the 'Analytical Models'. There are several strategies to create these models and one widely used is Machine Learning.
What is Machine Learning?
In the past, this type of methodology was something distant and more like science fiction, but today it is fundamental for business. Machine Learning is an area of Computer Science, created from research related to Artificial Intelligence. The increasing use of these methods is totally related to the strong computational growth. If in the past people tried to solve complex formulas almost manually, today technology is used to automate the construction of analytical models that use algorithms to learn from data in an interactive way. That is, the idea is that these algorithms are almost self-sufficient with as little human intervention as possible.
What makes Machine Learning different from Statistics or Statistical Models?
Machine Learning and Statistics are equivalent in some ways. This means that the results obtained with one methodology can be compared to another. However, when it comes to creating Statistical Models, the objective is to learn something related to the data, that is, to obtain insights from existing information. With Machine Learning, instead of simply trying to understand what happens in the data, what you have is the 'creation of examples' or 'rules', and with each execution of the algorithm, it is able to learn and improve from the examples. The interactive aspect of Machine Learning is important because models improve by learning from internal calculations producing better results.
Importance of Machine Learning Today
Machine Learning has historical importance. There are reports that this set of techniques was born during the Second World War and has been improved until the present day. To give you an idea, this methodology was already present, for example, in SAS since the 80's. At that time, through Machine Learning techniques, people were already looking to find behavior profiles in order to identify fraudsters.
The extraordinary impulse in the use of this methodology is due to a few reasons. First, because the amount of data is increasing at an unprecedented rate. For most organizations, the challenge is extracting valuable information from massive volumes of data from a wide variety of disparate sources. Another point is the storage options for this information – currently more accessible or even free. Also, computational processing power has never been cheaper or more powerful.
This means that with the right data, the right technologies, and the right analytics, you can quickly produce models that can analyze a large amount of data, regardless of its complexity, and deliver results in less time, more accurately, and with the minimal human intervention. The result? High-value forecasts that can guide better decisions.
Applications for Machine Learning are numerous, such as: fraud detection; online offer recommendations; real-time web and mobile advertisements; sentiment analysis based on textual sources from social networks; credit scoring; equipment failure prediction; new pricing models; network intrusion detection; writing pattern analysis; among others.
We can say that the central idea is to use a high performance technology, with an interactive aspect. What makes the use of Machine Learning advantageous is the possibility of creating systems that learn from their own data, obtaining the best results with minimal human intervention, generating highly effective results and enabling strategic decisions, even in real time.
EN 
PT 
