Intro to Feature Engineering

Imagine a retail business has a dataset that includes customer information such as age, gender, income, and the number of products purchased. Their goal is to build a model that can accurately predict whether a customer will make a high-value purchase.

At first glance, the dataset may seem straightforward, but feature engineering can help improve the accuracy of the predictive model by creating new features such as total spending, income category, and age group. But what exactly do we mean by features?

What are features?

To better understand this, let's consider an example of a dataset containing information about weather. Each record in the dataset may have various features that describe different aspects of the weather, such as longitude, latitude, humidity, wind direction, and atmospheric pressure. Hence, these characteristics of each weather records are referred to as the weather features.

Features are like building blocks that, when combined, provide a comprehensive picture of the data. However, the features in a dataset largely depends on the type of information in the data. Hence, the question now is what makes a good feature in a dataset?

Practice

Now that you have an idea of what makes a good feature in dataset, take a look at this COVID-19 dataset and mention TWO features that are NOT relevant to the analysis of COVID-19 cases according to the information provided.

Try to first work on this without checking the answer

Feature Engineering

As an analogy, feature engineering is like a craftsman shaping and refining the materials before building something. It's similar to a painter adding brushstrokes or a sculptor chiseling away to create a masterpiece. By carefully selecting, transforming, and creating features, data scientists can enhance the predictive power and understanding of their models.

For predictive models, the more features in a dataset, the more information and patterns to learn from the data, thereby improving the the model's predictive power. Let's look at some examples.

Example 1

Imagine you're a shop owner and a person comes up to you and asks for a tobacco. You proceed to ask for ID and you see the person's birthday is 09/12/1998.

This information is not inherently meaningful, but you add up the number of years by doing some quick mental math and find out the person is 25 years old (which is above the legal smoking age). What happened there? You took a piece of information 09/12/1998 and transformed it to become another variable age to solve the question you had - Is this person allowed to smoke? - that is feature engineering!

Example 2

Imagine a company that offers a streaming music service with a subscription model. They have a dataset containing information about their customers, such as age, gender, location, subscription duration, and the number of songs listened to per day. The company wants to improve their marketing efforts by identifying potential customers who are more likely to renew their subscriptions.

By leveraging feature engineering techniques, the company can create new features or modify existing ones to capture important customer behaviors and characteristics, such as creating features that capture engagement levels, playlist diversity, and usage patterns.

• Engagement Level: by combining the number of songs listened to per day and the subscription duration.
• Playlist Diversity: by creating a feature that measures the diversity of music genres in a customer's playlist.
• Usage Patterns: generating features that capture usage patterns, such as peak listening times or average session duration.

These engineered features help identify customers more likely to renew their subscriptions, consequently enabling targeted marketing campaigns.

➡️ In the next section, we'll be looking at Feature transformation 🎯.