Classification Problems

Name: Classification Problems
Author: Briana Brownell

Learn about the model building process and multi-class classification problems

Preview the first lesson free — get full access to all 6 lessons.

Course: On-Demand

In this chapter, we'll talk about classification problems in general, and give you some intuition on how to set up your classification modeling problem, and some basic considerations for the kinds of data that you'll need to solve these kinds of problems. Let's imagine that we want to create a simple classification model to determine whether a specific fruit or vegetable is an orange or not. So we have a list of different kinds of fruit, and we've looked at a couple different features that are probably going to be important in classifying whether or not the fruit or vegetable is in fact an orange. So we have three here, the color, the slice, shape, as well as the overall size. And by combining those three different features, we can determine whether or not the fruit or vegetable is in fact an orange. So you can imagine the classifier model would look something sort of like this flow chart. So we are initially looking at the color to determine whether or not it's orange, then looking at the vertical slice, and then finally the size in order to make that decision as to whether or not the fruit or vegetable is in fact an orange. This is an abstract representation of the way that a classifier model may work. So in this case, we have a flow chart, where it makes one decision, and then moves to the next one. That is not a necessary component for all of the models. Sometimes instead of a flow chart, you could have some of these decisions being made in parallel, but in order to represent what is happening in the classifier creation, it's easier to envision it as sort of a flow chart. Okay, so let's see how this works. So starting with a lime, we initially look at the color of the fruit or vegetable, and because it's green, we immediately know that it's not an orange. So what about for a carrot? So, because the color is orange, we're moving to that second stage, the shape of the vertical slice. In this case, since the carrot is pointed, we now know that the shape of the vertical slice is pointed or triangular, and so we know that that's not in fact in orange. Okay, but what about a pumpkin? This time, we have the color is orange. The shape of the vertical slice is rounded, but here, it fails at the last criteria, which is the size. So since it's large, our classifier says it's not an orange, which is of course, correct. But then when our classifier does in fact see an orange, all three of those attributes are correct, so we have the color's orange, the shape of the vertical slice is rounded, and the size is medium. And then our classifier will correctly say that the fruit in question is in fact an orange. So now you've seen a simple classification model, a classifier that will determine whether or not a fruit or vegetable is in fact an orange. And this classifier can be used as an example to model the way that any classifier might work.

Intermediate Provider : Briana Brownell 6 Lessons · 27m in Arabic, German, English, Spanish, French, Portuguese, Chinese

Course Description

These lessons will explain the typical model-building process in reference to the classification model pipeline. You will also receive an introduction to binary and multi-class classification problems, and the ways in which they are best used. You will learn how to use feature engineering to figure out the best data inputs for your classification models.

Additionally, these lessons will discuss and outline modeling with quality metrics. You will learn how to assess the quality of a classification model, and how to validate a classification model before its deployment.

What You'll Learn

Follow the typical process for building a classification model, from the model-building pipeline through deployment
Apply feature engineering to identify the best data inputs for classification models
Distinguish between binary and multi-class classification problems and when each is best used
Evaluate classification model performance using model quality metrics
Validate a classification model before deploying it
Build classification models for both binary and multiclass classification problems

Key Takeaways

The course explains the typical model-building process in reference to the classification model pipeline.
Feature engineering is used to figure out the best data inputs for classification models.
Binary and multi-class classification problems differ, and each has scenarios in which it is best used.
Quality metrics allow you to assess the quality of a classification model.
A classification model should be validated before its deployment.

Frequently Asked Questions

What does this course cover?

It covers the typical model-building process for the classification model pipeline, an introduction to binary and multi-class classification problems, feature engineering for choosing the best data inputs, modeling with quality metrics, and model validation methods.

What is the difference between binary and multi-class classification covered here?

The course introduces both binary and multi-class classification problems and discusses the ways in which each is best used.

How does the course address model quality and deployment readiness?

It discusses and outlines modeling with quality metrics, teaching how to assess the quality of a classification model and how to validate a classification model before its deployment.

What skills will I gain from this course?

You will build skills in Data Classification, Feature Engineering, Feature Learning, Feature Selection, Machine Learning Model Training, and Statistical Classification.

What topics are taught in the lessons?

The lessons cover the model building process, feature engineering, binary classification problems, model quality metrics, multiclass classification problems, and model validation methods.