Continuous Outlier Detection over Time-Series Data Streams
Kozue Ishida, Hiroyuki Kitagawa · 2008
Nowadays, we are facing with explosive increase of data, and data mining has become more and more important. Outlier detection is one of data mining issues and discovers outliers which have features different greatly from other objects or values. It is applied for fraud detection from network access and business data, abnormality detection from medical data and so on. With the recent diffusion of sensor devices and developments of network technology, we can continuously get a large volume of time series data from the real world. For outlier detection over time-series data streams, real-time response is very important. Therefore, we need efficient data processing. This paper proposes a continuous outlier detection method over time-series data streams, and shows its effectiveness.