Combined Approach to Detect Anomalies in Health Care Datasets

Abstract

Big data characterizing the quality of medical health care services provided to the population contain anomaly observations, which are either results of technical errors in filling databases, or falsified data. In the context of insurance medicine, the actual task is to identify such observation. The paper presents an approach to identify outliers in the database of medical health care information system based on the combination of Isolating Forest algorithm and subsequent neural network classifier.