Data science simplified: An introduction to classification

Merrian Webster defines classification as “a systematic arrangement in groups or categories according to established criteria.”

The world is full of classifiers. They help in preventing spam emails, identifying customers who may churn, predicting whether it will rain or not, etc. This supervised learning method is ubiquitous in business applications, but we often take it for granted.

In this post, I will discuss the key concepts of classification models.

Classification categories

Regression models estimate numerical variables and always see a target as a number. Classification models, on the other hand, have a qualitative target (aka categories).

In a large number of classification problems, the targets are designed to be binary, which implies that the target’s value will either be 0 or 1. These types of classifiers are called binary classifiers. Let’s take an example to understand this.

Let’s say a bank’s loan approval department wants to use machine learning to identify potential loan defaulters. In this case, the machine learning model will be a classification model. Based on what the model learns from the data it is given, it will classify the loan applicants into binary buckets:

• Bucket 1: Potential defaulters
• Bucket 2: Potential non-defaulters

The target, in this case, will be an attribute like “will_default_flag” and will be applicable for each loan applicant. It will take a value of either 0 or 1. If the model predicts it to be 1, it means that the applicant is likely to default on the loan. If it is 0, the applicant is likely not to default. Some classifiers can also classify the input into many buckets. These classifiers are called multi-class classifiers.

Linear and non-linear classifiers

Let’s say that we want to build a classifier that classifies potential loan defaulters. We’ll need to identify the features that determine potential defaulters first.

Let’s use the following features for this example:

• Income
• Credit rating

Confusion matrix

Conclusion