Detection of Outliers and Extreme Events of Ground Level Particulate Matter Using DBSCAN Algorithm with Local Parameters

Abstract

The critical negative effects of the particulate matter (PM) on human health are proven and hence the studies on the subject are increasing. Besides the health studies vast majority of the researches on particulate matter levels focuses on future projection and forecasting of the particulate matter concentrations. The data includes considerable amount of abnormal measurements. To perform an eligible analysis and prediction, a proper outlier analysis process is essential. However the studies focused on outlier identification in PM data are relatively few. This paper focuses on finding outliers and extreme events in ground level PM10 (particles smaller than or equal to 10 mu m in diameter) data using Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. The results show the effectiveness of the method to identify noise and extreme events.