1. Prediction models (unsupervised learning)
Great! Now we will review the topic of unsupervised learning models.
2. What is unsupervised machine learning?
Unsupervised learning models don't have a target variable and they are used to do clustering. The observations are grouped in clusters of similar elements. For example, this is broadly applied in customer segmentation.
Another use case is to detect anomalies that fall outside of discovered regular patterns or clusters, and then use that information either as input into supervised models, or as an input to decision making processes.
Finally, unsupervised learning models are used for recommender engines, for example recommending movies in Netflix or product recommendations on Amazon.
3. Clustering example - segmentation
We'll now take a slightly adjusted customer dataset with three variables - annual spend, recency in days - which is the number of days since customer's last purchase, and store visits in the last 12 months.
4. Segmentation - data
These variables are also called monetary value, recency and frequency.
5. Segmentation - training
Then we use these data points and the algorithm discovers similar groups or clusters.
6. Segmentation - discover
Finally, these clusters can be analyzed by looking at their average values in the original dataset.
7. Segmentation - analyze
Here we have discovered three segments. The first one displays medium spend, over two months since the last purchase, and 74 store visits last year on average. The second one has the highest spend, has been shopping very recently, and has a medium shopping frequency on average. Finally, the third segment spends the least on average, has been shopping pretty recently, and the frequency of shopping is pretty low.
8. Let's practice!
Great progress! Let's do some exercises to make sure we have learned the fundamental elements of unsupervised learning.