Get startedGet started for free

Why use ML for marketing? Strategies and use cases

1. Why use ML for marketing? Strategies and use cases

Hi there! My name is Karolis and I lead Analytics & Science team in Amazon. Welcome to the course of Machine Learning for Marketing! In this section we will review examples of how machine learning is applied in optimizing marketing strategy, learn typical data formats, and share insights into the best practices.

2. Types of machine learning

Let's start with describing three different types of machine learning. First one is supervised learning. These models use data about observations to predict a target variable. There are two kinds of supervised learning - classification and regression. In classification we attempt to predict a categorical variable, or a class. In the next chapter, we will build a classification model predicting whether a customer will churn. In the third chapter, we will build a regression model to predict customer purchases in the next month. The second machine learning type is unsupervised learning. Here, there is no target variable, and the models use different data points about observations to group them into similar clusters. A popular use case is customer segmentation. We will segment customers based on their product purchase history in the last chapter of this course. Finally, there is reinforcement learning which is outside of the scope of this course - the models in this space have agents that act on their own to maximize rewards defined by the environment. They are used in robotics and other advanced fields.

3. Supervised learning data parts and steps

Supervised learning models require two key data elements. The first one is the target variable which is what we want to predict. It could be predicting which customers will churn, or which customers will buy again. Another example is predicting how much the customers will spend in the next 30 days. The second data element are the features that will be used to predict the target variable. It could be purchase patterns and behaviors collected prior to customers churning, or a number of missed loan payments prior to defaulting on a loan.

4. Supervised learning data format

The data format for supervised learning model is typically an X by N plus 1 matrix with X number of observations - that could be customers, vendors, or products, and with N plus 1 number of columns - of which N is the number of independent features, and the last one - the dependent target variable that we will try to predict - for example whether a customer churned.

5. Unsupervised learning

In contrast, unsupervised learning techniques focus on using the data about the observations to group them into clusters with common characteristics. A popular example is customer segmentation based on the product purchase history.

6. Unsupervised learning data format

Unsupervised learning data is structured as an X by N matrix with X number of observations that we want to group together and N number of features. There is **no** target variable in unsupervised learning.

7. Let's practice!

Now, let's test your knowledge and explore some of the concepts we've learned!