Isolation forests: looking beyond tree depth
This work offers an incremental improvement to outlier detection methods, potentially benefiting data analysis applications that handle mixed data types.
The paper tackles the problem of outlier detection by improving the isolation forest algorithm's scoring method, showing that incorporating feature space size and point counts enhances performance, especially with categorical features.
The isolation forest algorithm for outlier detection exploits a simple yet effective observation: if taking some multivariate data and making uniformly random cuts across the feature space recursively, it will take fewer such random cuts for an outlier to be left alone in a given subspace as compared to regular observations. The original idea proposed an outlier score based on the tree depth (number of random cuts) required for isolation, but experiments here show that using information about the size of the feature space taken and the number of points assigned to it can result in improved results in many situations without any modification to the tree structure, especially in the presence of categorical features.