From the course: Data Preparation, Feature Engineering, and Augmentation for AI Models

Unlock this course with a free trial

Join today to access over 25,500 courses taught by industry experts.

Feature engineering: Scaling and normalizing data

Feature engineering: Scaling and normalizing data

From the course: Data Preparation, Feature Engineering, and Augmentation for AI Models

Start my 1-month free trial Buy for my team

Feature engineering: Scaling and normalizing data

“

- [Instructor] Sometimes we need to scale and normalize our data. This is because features with different ranges can distort model training and the predictions those models make. Now, many machine learning algorithms rely on distance calculations, so that requires that we work with comparable scales. And this is because we want to prevent large-magnitude features from really dominating the decisions that are made. Consider, for example, house sales. Now, the price of a house sale might be in the order of hundreds of thousands of dollars. That's one feature. Other features might be the number of bedrooms, which might be three, four, five, or so on. So the scales can really vary widely between different features. In addition, some algorithms will converge faster when we work with standardized and normalized data. And also, regularization techniques, which we use to prevent overfitting, work better when features are scaled. There are several coming scaling techniques. Min-max scaling…

Contents