This is the third article in the series entitled “What is AI?” If you have not already read the previous articles, we recommend that you do so. To access the first article, click here.
Let us illustrate Unsupervised Learning with an example. Imagine a child playing with different-coloured Lego bricks. The child can see that some bricks resemble others, and accordingly sorts the bricks into separate piles. The child has thus all by itself found a pattern and made a type of classification of the Lego bricks. In the field of ML this is known as unsupervised learning.
This method is highly useful if data is unstructured and/or complex. The result is patterns and classification models, often in the form of clusters. After that it is essential to manually interpret these clusters in order to link them to insights and/or actions. See the image below for a graphic illustration: an animation showing how an algorithm known as “k-means clustering” functions can be seen in the following video.
One common application area is data-driven segmentation of customer bases into different customer groups. These clusters of customers can then be interpreted as certain categories of customers who should be treated in a particular way. Such segmentation can then be used for marketing purposes, for example in a “marketing-automation tool”, which can direct differentiated campaigns in a better way.
Another common example is to identify anomalies (things that stand out). Banks often apply this approach on their transactions to identify potential fraud. They do this in order to improve the customer’s experience and to reduce costs in the form of paybacks. More or less the same approach can be applied for analysis of sensor data, for instance in a production line to quickly identify defective products.