Logistic regression is an algorithm that deals with classification issues and problems. It's one of the best known and used machine learning algorithms in the world, being used in different areas, such as cybersecurity and biology. First of all, it’s important to keep in mind that logistic regression isn’t used only to classify things between two categories. For example, if an email is spam or not, or if a person sick or healthy. Yes, logistic regression is widely used for two-class problems. But not only for this.
Being a little more technical, logistic regression works with concepts of statistics and probability. As appropriately stated by Wikipedia, “logistic regression measures the relationship between the categorical dependent variable and one or more independent variables by estimating probabilities using a logistic function”. This means that this type of machine learning algorithm analyzes different aspects or variables of an object to later determine a class in which it fits better.
Types of logistic regression
As we've already explained, logistic regression is widely used to categorize objects between two classes. But it's not restricted to that. In all, there are three main types of logistic regression. Or we can say that there are three main models. Let's take a look at them.
1. Binominal logistic regression
In the binomial logistic regression model, objects are classified into only two groups or categories. It's almost a game between what is and what isn't. For example, the message is spam or not, the image is colored or not, the cell is cancerous or not.
2. Ordinal logistic regression
The ordinal logistic regression model is different because it works with the concept of ordered categories. In this case, objects are classified into three or more classes that have a determined order. For example, the athlete's performance is poor, fair or excellent. Another example: the level of patient satisfaction with the treatment is dissatisfied, satisfied or highly satisfied.
3. Multinomial logistic regression
In the multinomial logistic regression model, objects are classified into three or more categories that have no order among them. Let's go to the examples. This animal is a cat, a lion or a tiger. This fruit is an apple, a pear, a mango or a passion fruit.
Using logistic regression in emails
Logistic regression is a very effective type of machine learning algorithm. That's why it's one of our development's team favorite algorithms. Consequently, it's one of the most used algorithms in our email security and protection solutions. Here, at Gatefy, we’ve adopted the multinomial logistic regression model as one of the main mechanisms of our artificial intelligence system.
To make it simpler, let's create an example using your business. Regardless of the size your business, it should receive tens, hundreds, or even thousands of emails daily. Now imagine that your email protection solution classifies messages into just two groups, using binomial logistic regression: spam (unwanted messages) and ham (desired messages). How would your company's employees inbox look like today?
It would probably be a little bit messed up by pointing out many false positives and false negatives. That is, a bunch of desired emails would be blocked and a bunch of unwanted emails would be delivered. The result is that employees would waste a lot of time analyzing these emails and, worse, they would be more exposed to frauds that can put the entire business at risk, such as phishing attacks.
In our email security solution, Gatefy’s artificial intelligence with the help of multinomial logistic regression allows organizations to classify their emails into at least seven different classes. The result: more security and better email management. That is, the organization has more visibility and control over the information and still reduces the risk of suffering a data breach. In addition, employees can focus on the core business of the organization.
The cool thing is that Gatefy's artificial intelligence system will learn from the organization's email traffic. With each passing day, the solution becomes more assertive and accurate, improving its performance. As someone always says to me, it's technology being used for good and for you.