Parallel Algorithm of Flow Data Anomaly Detection Based on Isolated Forest

The isolated forest algorithm is improved and applied to the hydrological field. The parallel anomaly detection algorithm (Flink-iForest) is proposed. At the same time, the k-means algorithm is combined to solve the problem of Flink-iForest threshold division and improve the stability of anomaly detection results. Through various experiments and real hydrological data, first of all, the Flink-iForest algorithm is verified in terms of accuracy, efficiency and scalability, and compared with the standard SKlearn-iForest and PIFH algorithm; finally, the effectiveness and efficiency of Flink-iForest algorithm are proved by experiments.