isolation-forest

Leverages Spark's ML library for scalable training and scoring on large datasets, with support for pipelines and HDFS persistence, as shown in the distributed training examples.

Includes Extended Isolation Forest with random hyperplane splits to eliminate axis-aligned bias, improving detection in correlated feature spaces, evidenced by the heatmap comparisons and benchmarks.

Supports ONNX export for standard Isolation Forest models, enabling portable inference across languages and platforms via the Python-based converter detailed in the README.

Inherits from Spark's Estimator and Model classes for seamless integration into existing ML workflows, with model persistence and parameter tuning features.