Continuous outlier detection on uncertain data streams

Abstract

Time series data streams are common due to the increasing usage of wireless sensor networks. Such data are often accompanied with uncertainty due to the limitations of data collection equipment. Outlier detection on uncertain static data is a challenging research problem in data mining. Moreover, the continuous arrival of data makes it more challenging. Hence, in this paper, the problem of outlier detection on uncertain time series data streams is studied. In particular, we propose a continuous distance-based outlier detection approach on a set of uncertain objects' states that are originated synchronously from a group of data sources (e.g., sensors in WSN). A set of objects' states at a timestamp is called a state set. Generally, the duration between two consecutive timestamps is very short and the state of all the objects may not change much in this duration. Therefore, we propose an incremental approach of outlier detection, which makes use of the results obtained from the previous state set to efficiently detect outliers in the current state set. In addition, an approximate incremental outlier detection approach is proposed to further reduce the cost of incremental outlier detection. Finally, an extensive empirical study on synthetic and real datasets is presented, which shows the efficiency of the proposed approaches.