This post will give you a brief explanation of what Machine Learning is, what role it plays in our society, and how it works.

What is a Machine Learning model and how does it work

I believe there is no better way of describing Machine Learning, what it is, and does, then starting with a picture.

illustrating how a machine learning model learns from data
How a model learns from data (very simplified)

The above image tries to show how a model learns from data. However, more importantly, the image attempts to illustrate what often is necessary to do as a human before a Machine Learning Algorithm can learn.

Let’s go through the image from left to right and talk about each step.

Input Raw Data

We humans learn from experience. It is no different for a Machine Learning Algorithm. The only difference is for algorithm data is the experience. Therefore, without data, we cannot train a Machine Learning Algorithm. To put it harshly without data there is no Machine Learning.

The human interaction (Supervisor, Trining Data set, Desired Output)

Before an algorithm (model) can be trained more often than not the raw data must be prepare and labeled. Labeling is the process of taking each data point and assigning it an explanation. For example, if you want to train a model to take an email as an input, look at it and give you the output, indicating if the email is fraud or not, then the model needs to be trained on many examples (input raw data), where you show the algorithm both emails that are fraudulent and other emails that are not. But how should the algorithm know which is which while looking at the training data, for that purpose a human often has to take each email and put a label on it saying, for example, “fraud” or “no fraud”.

How does the algorithm learn from experience (Algorithm)

When the raw data is prepared and labeled then the algorithm is able to look at thousands of emails, look at the text and the label. The algorithm will then learn the patterns of the text in emails associated with the label “fraud” and the patterns of the texts associated with the label “no fraud”.

Asking the model a new question (Processing & Output)

When a Machine Learning model is trained and ready to be used on new, unknown data. In the example with emails, new data is simply emails that the model has never seen. If we now give the model a new email to analyze, then the model will compare the pattern in the new email with the learned patterns of all the emails that were labeled “fraud” and all the emails that were labeled “not fraud”. If the pattern in the new email has more in common with the learned pattern from the “fraud” group then the response of the model will be that with some likelihood the new email might be fraudulent.

Now we have a fair grasp of what Machine Learning is and even what it takes to train one. There are of course many more tasks involved in training a model, however, that explanation is out of scope for this post. Instead, let’s discuss what role Machine Learning plays in our society.

Download the free Machine Learning project checklist.
8 steps that will guide you through your project.

What is Machine Learning used for in our society

Again there is no better way of explaining a topic then with a picture.

picture of a city full of machine learning
City full of Machine Learning

Machine Learning (ML) is kinda everywhere nowadays. You might not realize it but you are probably interacting with a model multiple times each day.

Here are some examples, and these are probably the more obvious ones, there are a lot more out there. The hard part is it is becoming more and more difficult knowing when we are interacting with a machine learning model.

ML in E-commerce

Online stores do not have the possibility to talk to a customer in the same way as a salesperson can do in a physical store. This makes learning the customer to know difficult. Instead of online stores attempt to gather information about a potential customer by looking at the interactions the customer has made while on the web-page. ML models can help to make sense of the data and even further use all the gathered data to train a model in making customer segmentation to present more customized marketing to each visiting customer. Here are even more use-cases on how ML can be used in e-commerce.

ML in financial institutions

Like in e-commerce, financial institutions use ML to perform customer segmentation. With customer segmentation, a bank can predict what type of products a customer might need in the future. Or even predicting if a customer is more or less prepared to leave the bank.

Financial institutions also use ML for process optimization and automation. Banks receive a lot of documents from customers. Often the documents need to be inspected, approved and archived. ML models look through all the documents and automate the process of inspection and approval. Let’s not forget trading. ML is most certainly also used for trading stock options. You can read more about ML in banking here.

ML in cars

Now-days it is almost not possible to buy a car that does not use ML to help us navigate the car.

When we think about ML in cars we mostly think about self-driving cars. However, there are a lot more examples. For instance, cars use ML for parking assist, lane-change assists, and smart energy systems. There are probably a lot more examples which I don’t know about. You can read more about ML in the automotive industry here.

The list of real-world examples where Machine Learning is used in our society can probably continue for a while. Even thou a longer post would help my SEO I will stop here. While we are at the SEO topic, Google most certainly uses Machine Learning when you use their search engine. They are also using ML for validating and classifying web-pages and their content.

If you want to start learning about Machine Learning and start working in that area. Head over to my comprehensive yet manageable course Introduction to Machine Learning End-to-End. This course will take you end-to-end through the process of working on a real-world Machine Learning project. At the end of the course, you will have a complete code example that you can use for other projects.