1. Welcome and Introduction
Hi. This is Nina Zumel and John Mount from Win-Vector LLC. Welcome to our course on regression in machine learning. Regression is the task of
2. What is Regression?
predicting a numerical outcome based on the values of a set of inputs, or independent variables.
In the statistical sense, regression is predicting the expected value of an outcome, but in the casual sense, and for the purposes of this course, we'll define regression as predicting numerical values.
This distinguishes regression from classification, which is the task of making discrete predictions.
3. What is Regression?
Predicting how many units of a product will sell, or how much it will sell for, are regression problems.
Predicting if a customer will buy a product (yes or no) is a classification problem.
Let's walk through an example regression task.
4. Example: Predict Temperature from Chirp Rate
Here we see a scatterplot of how fast a cricket chirps (chirps per second) on the x axis, and the temperature on the y axis. Let's say our goal is to predict the temperature by measuring cricket chirp rate. Suppose we fit a linear
5. Predict Temperature from Chirp Rate
regression model to this data (we'll show you how later in this course). The predictions from this model are shown by the blue line. If we assume that temperature is linearly related to cricket chirp rate, then the blue line is the best fit line through the data, and it predicts what the expected temperature should be, based on the observed chirp rate.
6. Predict Temperature from Chirp Rate
For example, suppose we hear a cricket chirping at 16.5 chirps/second. We see here that this model predicts a temperature of about 80 degrees.
7. Regression from a Machine Learning Perspective
Let's step back and look at the reasons for modeling. Modeling from a scientific mindset focuses on understanding the process that produced the data: how each variable affects the outcome.
Modeling from an engineering or machine learning mindset focuses on predicting future events accurately, and less on the relationship between variables and outcomes.
In this course, we will emphasize predicting accurately. You will learn several algorithms for fitting regression models. By the end of this course, you should have a better idea about the advantages and disadvantages of each algorithm.
Now let's do an
8. Let's practice!
exercise to quickly review what you've learned.