Machine Learning

What exactly is machine learning?

Machine learning is a subset of artificial intelligence (AI) that deals with the extracting of patterns from data, and then uses those patterns to enable algorithms to improve themselves with experience. This type of learning can be used to help computers recognize patterns and associations in massive amounts of data, and make predictions and forecasts based on its findings.

A good example is teaching a computer to play a game of Go. A computer can be taught the rules of the game in such a way that it can adapt and respond to a limitless number of moves, including those it has never encountered. In 2016, for the first time ever, a machine beat the world’s best “Go” player because of machine learning.

Machine learning is evolving rapidly. Although some forms of machine learning have arguably been around for hundreds of years, it’s now front and center in the world of technological innovation. Today, it can be used in nearly any field or industry to consume massive amounts of data from unlimited sources and drive real business impact.

The evolution of machine learning

Although artificial intelligence is nothing new (the famous Turing Test was developed in 1950, and AI was established as field of research in 1956), it’s grown to include common and impactful applications in both business and everyday life. This progress is made possible, in large part, from machine learning.

Relatively simple forms of machine learning like linear regression algorithms have existed for 200 years. But modern machine learning represents a paradigm shift in how computing and statistical analysis was done in the past, where the emphasis was on rule-based decision making: give the computer a set of rules and deterministically have it make decisions.

In many cases, this would lead to the computer being correct. However, it couldn’t handle data outside the rules provided to it. If, for example, the world changed, leading to a change in the data that you were analyzing, the model wouldn’t know what to do with it.

Why is machine learning so important?

Humans may be smart, but we often just can’t see well enough. There’s a lot we might want to know about our business, but the patterns we need are hidden in dense data. Machine learning lets us teach a computer to look at the same data that we’re looking at, and to then derive patterns and connections that we can’t see. This provides truly superhuman insight into the huge mass of data being generated today, driving a revolution in nearly every sector of business.

Machine learning is already making a substantial difference in several industries. In the financial services sector, machine learning is being used to analyze data for the purpose of risk analytics, fraud detection, and portfolio management. In travel, with GPS traffic predictions. It’s also used to populate Amazon and Netflix recommendations.

The implications of this progress are huge, but in order to fully appreciate them, you need to understand more about how machine learning works.

Machine learning and artificial intelligence​

First and foremost, machine learning and artificial intelligence are not the same thing. Machine learning is only a subset of artificial intelligence. But, if that’s the case, what else is a part of this field? AI, machine learning and deep learning have specific relationships with each other and work together. The image below does a good job explaining this.

It’s important to avoid the misconception that machine learning and artificial intelligence are the same. They are not the same and the differences truly matter.

Artificial intelligence (AI) is the simulation of human intelligence in computers. It refers to any set of approaches or techniques, ranging from the pretty simple to the incredibly complex, that is employed to get a computer to mimic human behavior. If you create a set of simple rules that never allow a computer to lose at rock-paper-scissors, then you’ve created some basic AI.

As we’ve established above, machine learning is a subset of artificial intelligence. It is one of the approaches used to achieve AI. With ML, however, the focus is on teaching the computer to learn for itself how to complete a task.

Another term we hear mistakenly interchanged with machine learning and artificial intelligence is deep learning. Deep learning is a specific method for machine learning – not an entire field of its own.

How exactly does machine learning work?

Machine learning operates on the basis of algorithms that enable computers to find patterns in data and them turn them into optimal behavior.

Machine learning algorithms use computational techniques to “learn” information directly from data without relying on a predetermined equation as a model.

Machine learning algorithms can also adaptively enhance their performance as the new data becomes available—what we call model resiliency.

There are many techniques involved in machine learning, such as deep learning, support vector machines, decision trees, k-means clustering, and a host of others.

These techniques fall into two categories: supervised learning and unsupervised learning.

Common machine learning methods

What’s the difference between supervised and unsupervised machine learning? On the most basic level, the key difference is whether or not you tell your model what you want it to predict. Let’s dive into the details of each.

Supervised

Supervised learning algorithms are taught using labeled examples, such as an input where the desired output is known.

For instance, a piece of equipment with data points labeled either “F” (failed) or “R” (runs). The algorithm gets a set of inputs along with the corresponding correct outputs and learns by examining its actual output with the right outputs to detect errors, then it modifies the model accordingly.

By using systems like classification, regression, prediction, and gradient boosting, supervised learning utilizes patterns to predict the values of the label on other unlabeled data. Supervised learning is typically used in applications where historical data predict likely future results. For example, it can anticipate who is expected to file an insurance claim.

Unsupervised

Unsupervised learning is used with data that has no historical labels. This means the system is not given the “right answer.” The primary objective is to explore the data and find some structure within it.

Unsupervised learning works well on transactional data. It can identify groups of customers with similar attributes who could be treated similarly in marketing campaigns. Or, it can find the main characteristics that separate customer populations from one another.

The most well-known techniques include self-organizing maps, nearest-neighbor mapping, k-means clustering, and singular value decomposition. These algorithms are also used to segment text topics, recommend items and identify data outliers.

Some use cases don’t fall easily into either supervised or unsupervised learning categories. Particularly in cases where some or none of the data points have labels or output data, semi-supervised methods are used.

Functionally, semi-supervised learning is just a mix of supervised and unsupervised. For example, you may use unsupervised learning to categorize a collection of emails as spam or not spam. You could then use a supervised learning technique with information like the frequency or usage of certain words to determine whether it is spam or not.

Choosing the right machine learning solution

Machine learning involves a huge amount of data. The process of feeding in the data itself is historically tedious and requires a lot of manual coding. This limits the ability of organizations to effectively utilize the advantages of machine learning, causing delays that can make models outdated even before they are complete.

And it doesn’t end with just having to feed huge amounts of data into some machine learning model. In order to get the most value from ML, you need to pair the best algorithms with just the right tools and processes. That’s when the whole thing comes together.

RapidMiner provides an end-to-end data science platform that’s built to deliver business impact. It unifies data prep, machine learning and model operations to enhance the productivity of users of any skill level across an enterprise. With a cutting-edge solution like RapidMiner, insights can be delivered on a scale and speed greater than was ever possible.

RapidMiner AI Assessment Matrix

Find out how ML can transform your business

We apply our expertise to help you identify the use cases you should tackle in your organization. The outcome is an impact-feasibility map that you can use with or without us.

Additional Machine Learning Resources