Anomaly detection based on improved isolated forest

Anomaly detection is one of the hot research fields of machine learning and data mining, which is mainly used in fault diagnosis, intrusion detection, fraud detection and other fields. At present, there have been a lot of effective research work, especially the anomaly detection method based on isolated forest, but there are still many difficulties in processing high-dimensional data. A new anomaly detection algorithm k-nearest neighbor based isolation forest (KNIF) is proposed. In this method, the hypersphere is used as the isolation tool, the k-nearest neighbor method is used to construct the isolation forest, and the outlier calculation method based on distance is constructed. Sufficient experiments show that KNIF method can effectively detect anomalies in complex distribution environment and adapt to different distribution scenarios.