Machine learning has generated quite the buzz – from Elon Musk fearing the role of unregulated artificial intelligence in society, to Mark Zuckerberg having a view that contradicts Musk's.

So, what exactly is machine learning? Simply put, machine learning is a set of methods that can detect patterns in data and use those patterns to make future predictions. Machine learning has found immense value in a wide range of industries, ranging from finance to healthcare. This translates to a higher requirement of talent with the skill capital in the field of machine learning.

Broadly speaking, machine learning can be categorized into three main types:

• Supervised learning
• Unsupervised learning
• Reinforcement learning

Scikit-learn is designed to tackle problems pertaining to supervised and unsupervised learning only, and does not support reinforcement learning at present.

Supervised learning is a form of machine learning in which our data comes with a set of labels or a target variable that is numeric. These labels/categories usually belong to one feature/attribute, which is commonly known as the target variable. For instance, each row of your data could either belong to the category of Healthy or Not Healthy.

Given a set of features such as weight, blood sugar levels, and age, we can use the supervised machine learning algorithm to predict whether the person is healthy or not.

In the following simple mathematical expression, S is the supervised learning algorithm, X is the set of input features, such as weight and age, and Y is the target variable with the labels Healthy or Not Healthy:

Although supervised machine learning is the most common type of machine learning that is implemented with scikit-learn and in the industry, most datasets typically do not come with predefined labels. Unsupervised learning algorithms are first used to cluster data without labels into distinct groups to which we can then assign labels. This is discussed in detail in the following section.

Unsupervised learning is a form of machine learning in which the algorithm tries to detect/find patterns in data that do not have an outcome/target variable. In other words, we do not have data that comes with pre-existing labels. Thus, the algorithm will typically use a metric such as distance to group data together depending on how close they are to each other.

As discussed in the previous section, most of the data that you will encounter in the real world will not come with a set of predefined labels and, as such, will only have a set of input features without a target attribute.

In the following simple mathematical expression, U is the unsupervised learning algorithm, while X is a set of input features, such as weight and age:

Given this data, our objective is to create groups that could potentially be labeled as Healthy or Not Healthy. The unsupervised learning algorithm will use a metric such as distance in order to identify how close a set of points are to each other and how far apart two such groups are. The algorithm will then proceed to cluster these groups into two distinct groups, as illustrated in the following diagram: