Machine learning in Business

A field of study that gives computers the ability to learn without being explicitly programmed”. That is how Arthur Samuel defined machine learning by the end of the 1950s. As one of the pioneers in the field of Artificial Intelligence, he was the first to popularize this term as one of the subsets of AI. On the other hand, Machine Learning is a set of techniques used in the field of Data Science, for extracting insights.

Technically speaking, machine learning includes defining a mathematical model which represents the problem we want to analyze, in order to find an approximate solution.

In most cases, machine learning models are prediction-based: trained on some (hopefully representative) sample of data, they are learned to provide an output (used for decision making purposes, for example – the answer to the question “Is a particular customer going to churn?), based on some input features (for example, a set of attributes describing the customer, his preferences and activity level). This is the so-called supervised learning.

There is also unsupervised learning – where the output is not predefined, but extracted as insight from the algorithm (for example, in clustering we don’t know what the clusters are, prior to running the model, but the model itself identifies groups of similar instances, called clusters and returns the cluster labels as an output).

Besides these two groups, there are other variations on the topic – like semi-supervised and reinforcement learning, as hybrid and modified versions developed to tackle other types of problems.

The choice of the appropriate algorithm depends on several things: a problem being defined, data and its nature, model pre-assumptions, computational power and resources needed. Enough with the technicalities – what does machine learning mean to the business today?

Let me depict the ML application in one picture (a slight digression: all the credits go to my dear colleague, mr. Milos Milovanovic, for drawing this picture, making this easier for me).

This picture is a showcase of our everyday life. It gives a sneak peek on how the biggest leaders in the AI world are using machine learning. You wake up, commute to the work, eat, plan a trip, etc – without even noticing how every application that you use is driven by one or more machine learning algorithms developed to customize and personalize the content you are watching/using, in order to leverage your experience and satisfaction.

For example, Google integrates ML in all its applications, for natural language processing purposes (Google search engine, speech recognition and language translation), image processing (Google photos) and traffic prediction (Google maps).

Uber sets dynamic drive prices based on the  traffic jams and demand rates.

Porhub uses computer vision and recommender systems to place the right content for its visitors.

Netflix is also using collaborative filtering recommender systems for tailor-made recommendations to the users.

Facebook is one of the biggest leaders in AI research, integrating ML in all its services, while simultaneously open-sourcing some of the algorithms (like Facebook Prophet, used for forecasting).

More and more companies are becoming aware of the importance and benefits of ML applications within the business. Applications often include:  extracting predictions of the future – classification, regression and forecasting; and extracting insight explaining underlying patterns, similarities and drivers of some events that happened. Having this as a tool in the decision-making process can be a really powerful asset that could affect the perception of the business and its future.

However, machine learning strongly relies on the experience. It requires lots of historical data and lots of observations – in order to catch all the possible patterns. Machine learning is only as strong as the prediction power of the data is (if the dataset is not representative, or there are not enough observations, the model will most possibly fail).

We already mentioned that ML is a subset of one much broader discipline – Artificial Intelligence. Artificial Intelligence is a branch of computer science that deals with (ML) algorithms inspired by various facets of natural intelligence. It is a system able to completely autonomously perform tasks that normally require human intelligence, like visual perception, speech recognition, problem solving, language translation,… Besides being autonomous, it is able to make decisions and define actions in dynamic environments, where the conditions and constraints are changing (which is not the case with traditional Machine Learning, strongly relying on the experience and already seen patterns). The next picture represents the level of intelligence we’ve artificially reached so far.

However, as much as we talk about it nowadays, we are extremely far from the AI at its finest. Thus, in most cases, when companies talk about AI, they usually refer to the application on machine learning and switching to data-driven decision making (a.k.a. Data Science).

I would like to conclude this post with one quote authored by one of the biggest businessmen in the world – Ray Dalio.

“If the future can be different from the past and you don’t have a deep understanding, you should not rely on AI.”

Machine learning can be powerful, but only when wisely taken and deeply thought through. Being aware of this could make a significant difference between success and failure of integrating machine learning in business.

About Valentina Đorđević 12 Articles
Valentina Djordjevic is a huge Data Science enthusiast. Graduating from the Faculty of Organizational Sciences, she faced the challenge of packing her (then irreconcilable) interests into a unique career path-cutting machine. Still, she is a real "fox" when it comes to her areas of interest, so we have no doubt that this has given her good signposts. Valentina finds motivation in challenging projects and in working with equally enthusiastic team members. She believes that working on the Data Science challenges helps her to continually grow, examine (her) boundaries, and develop her imagination. Still, she has not yet determined whether the fact that she has never solved the same problem in exactly the same way is a good or bad thing. She hopes her colleagues, or maybe someone in the community, will help her finally come up with the correct answer.

Be the first to comment

Leave a Reply

Your email address will not be published.