April 2018
Beginner to intermediate
282 pages
6h 52m
English
A multivariate outlier is a blend of extreme scores on at least two variables. Univariate outlier detection methods are suited well to dealing with single-dimension data, but when we get past single dimension, it becomes challenging to detect outliers using those methods. Multivariate outlier detection methods are also a form of anomaly detection methods. Techniques such as one class SVM, Local Outlier Factor (LOF), and IsolationForest are useful ways to detect multivariate outliers.
We describe multivariate outlier detection on the HR attrition dataset using the following IsolationForest code. We need to import the IsolationForest from the sklearn.ensemble package. Next, we load the data, transform ...