What is machine learning? If you have an interest in analytics, you have probably heard this term a lot. Many people use it incorrectly, and some predict grand outcomes for its future. Despite the hype, machine learning is one of the most powerful technologies of the modern enterprise. You will soon see why.
What is machine learning?
Machine learning is a school of computer science focused on programming machines to improve their own performance through data and iteration. In traditional programming, a human engineer needs to code every process the machine carries out, which makes some types of problem hard to solve.
Take the example of a video streaming service such as Netflix. To make recommendations, engineers have to decide which videos to recommend based on user feedback: it is an “if A, then B” case. From one point of view, this works. But how can programmers do this for thousands of titles and millions of unique user histories? And can they really know that someone who enjoys one title will enjoy another?
Machine learning solves these problems. Instead of relying solely on human instruction, it uses algorithms to gather data, learn from it and make predictions. The system can adjust as users provide more information. This often leads to recommendations a human would not predict.
AI vs. machine learning
One of the most common questions is the difference between machine learning and artificial intelligence. The answer is simple: machine learning is a type of AI. It is a subset within the broader field, in the same way that artificial intelligence is a subset within the broader field of computer science. Unlike machine learning, AI is a rather non-technical term: it is more about outcomes than about a specific methodology. Any intelligent system that mimics human behaviour is AI.
Supervised learning vs. unsupervised learning
Supervised learning is a common task that works by using input-output pairs to train an algorithm. Let’s use the example of a team that wants to train an application to recognise images of dogs. Dogs can look very different from one another, and it is extremely hard to teach a computer to recognise the similarities between a Great Dane and a Chihuahua. Any traditional rule would have exceptions, which is why image recognition is a good use case for machine learning.
Instead of programming the machine to look for certain features, engineers build a program that iterates based on the input and output pairs. Humans provide a large number of images as input, each with an output indicating whether or not it includes a dog. With enough information and training, this iterative model can achieve surprising accuracy and will no longer require human training.
In unsupervised learning, the system iterates without the labelled, structured data used to train supervised algorithms. The algorithm has to train on inputs alone, without knowing the corresponding outputs, and tries to model the underlying structure of the data in order to understand it and predict outcomes.
Why should business leaders care about machine learning?
This technology has already dramatically altered the business landscape. Take e-commerce: the moment you open your account with a large online retailer, you receive highly accurate recommendations that benefit both seller and buyer. Could a traditional algorithm make recommendations this accurate and personalised for millions of customers?
Other use cases include cybersecurity, healthcare, process automation and financial analysis. There is a strong chance this technology has already affected your operations. The question now is how your team will use it to succeed.
The natural next step for machine learning in business is agentic analytics: AI agents that answer questions about your governed data. At Digital Fox Data we build that foundation with our Business Intelligence consulting in Qlik, Tableau and Power BI. Want to apply AI to your data with confidence? Let’s talk.