课程: TensorFlow: Neural Networks and Working with Tables

今天就学习课程吧！

今天就开通帐号，24,700 门业界名师课程任您挑！

Feature engineering

Feature engineering - TensorFlow教程

课程: TensorFlow: Neural Networks and Working with Tables

开始一个月试用 购买团队方案

Feature engineering

“

- [Instructor] In a nutshell, feature engineering is using our domain knowledge to extract features from the raw data. So let's see what we can do with the numeric and categorical columns. You can see that there are a couple of numeric columns, such as age, the fare paid, the number of siblings and spouses that traveled together, and the number of parents and children. Now in general, machine learning algorithms don't perform well when the input numerical attributes have different scales. Let's take a look at a range of values using describe. Now you can see that the minimum values for most of the numeric fields are close to zero. The maximum values vary a fair bit where we have 512 as the maximum value for the fare paid, but for the number of siblings and spouses, the maximum number is only eight. Now that's quite a big difference, which is why we apply normalization. The result is the mean of each feature is zero…

内容