April 2018
Beginner to intermediate
282 pages
6h 52m
English
As the names imply, univariate outliers are based on a single attribute in a dataset. Univariate outliers are discovered using box plots and by seeing the distribution of the values of an attribute. However, when we build AutoML pipelines, we don't have the privilege to visualize the data distribution. Instead, the AutoML system should be able to detect the outliers and treat them by itself.
So, we can deploy any of the following three methods for automated univariate outlier detection and treatment:
Let's create a dummy outlier dataset to demonstrate the outlier detection and treatment method:
%matplotlib inlineimport numpy as npimport matplotlib.pyplot ...