Data Anomaly Detection with Parallelizing CDP Algorithm

Many existing data is inconsistent due to various reasons, such as noise, missing data, or repeated errors. Previously, the CDP clustering algorithm has been shown as an effective method to detect data anomalies. Yet, single computational processor or sequential version of the CDP algorithm demands lengthy computational times and often not acceptable in practice when dealing with voluminous data. In this paper, we propose the PCDP algorithm on detecting data anomalies by parallelizing the CDP algorithm with the support of multiple machines. Through preliminary experiments, the PCDP algorithm has demonstrated its fast computation performance as well as retaining the anomaly detection accuracy for customer data of a commercial bank.