Unsupervised Fraud Detection in Environmental Time Series Data

Time series often contains outliers and structural changes. These unexpected events are of the utmost importance in fraud detection as they may pinpoint suspicious activities. The presence of such unusual activities can easily mislead conventional time series analysis and yield erroneous conclusions. Traditionally, time series data are first divided into small chunks. kNN-based outlier detection approaches are then applied for monitoring behaviors over time in data mining. However, time series data are very large in size and cannot be scanned multiple times, and as they are produced continuously, new data are arriving. To cope with the speed they are coming, in this chapter, we propose a simple statistical parameter-based anomaly method for fraud detection in environmental time series data. The results of the experiments performed show that the proposed algorithm is effective and efficient .