Learn about some common uses of regression.
- [Instructor] Value prediction is an incredibly useful technique because we can use it to solve so many different kinds of problems. In this course, our project is to build a system that can predict the value of a house based on it's attributes. But you can use the exact same technique to value any of kind of product as long as you have training data. For example, we could build a system that can estimate the value of a used car based on it's attributes such as it's age, it's condition, color, and mileage. Or we can build a system to estimate the value of used cell phones that we want to resell based on the phone's features and past sales of similar phones.
Another use of value prediction that affects your daily life is fraud detection. When you buy something online or use a credit card there's a very good chance your purchases run through a fraud detection algorithm. In this case, the model uses details about your purchase to decide how likely your purchase is to be fraudulent. If the model returns a high likelihood of fraud the purchase will be blocked. Likewise, value prediction is used to model the risk of issuing a home loan. In this case, the inputs are details about the borrowers, and the output is how likely the loan is to be paid back. This helps banks decide which loans are worth the risk.
But value predictions isn't limited to financial transactions. For example, you can build a model where the inputs are the words that appeared in the movie review and the output is how positive or negative the review is. This is called sentiment detection. It allows a computer to look at a piece of text written by a human and guess if the human was writing a positive or negative review. There are also many uses for value prediction in the medical field. Machine learning models are often used to help doctors read x-rays and other types of medical images. These models are sometimes more reliable than doctors at detecting diseases.
AI researchers have even predicted that within five years computers will be more reliable overall than radiologists at reading x-rays. The reason value prediction algorithms are so useful is because so many problems can be modeled as value prediction problems. What if the input to the algorithm is a picture taken with a video camera and the output was the angle to turn a steering wheel and the amount to press the gas pedal? You've just created a simple self-driving car! Learning how to model the problem as a value prediction problem is a very useful skill. Once you figure out the inputs and outputs you can usually build a solution with machine learning.
- Setting up the development environment
- Building a simple home value estimator
- Finding the best weights automatically
- Working with large data sets efficiently
- Training a supervised machine learning model
- Exploring a home value data set
- Deciding how much data is needed
- Preparing the features
- Training the value estimator
- Measuring accuracy with mean absolute error
- Improving a system
- Using the machine learning model to make predictions