Time series data streams are common in wireless sensor networks in nowadays. This type of data is having uncertainty due to the limitation of the measuring equipments or other sources of corrupting noise, leading to uncertain data. As uncertain streaming data is continuously generated, mining algorithms should be able to analyze the uncertain data. To detect the outliers in this project we propose two continuous distance-based outlier detection approaches (an exact and an approximate) are proposed for uncertain time series data streams. These two algorithms are implemented based on the cell based approach. These two approaches can be applied on uncertain objects. A set of uncertain objects at particular time stamp is called state set. As the duration between the two time stamps is very less to detect the outliers we use the incremental approach (use the results obtained from the previous state set to detect outliers in the current state set). An approximate incremental outlier detection approach is proposed to further reduce the cost of incremental outlier detection. Cell based algorithm is employed for the efficient detection of outliers within a state set, in both the incremental algorithms. To show the efficiency of the proposed approaches synthetic and real datasets are used.

Clustering Outlier detection Cell-Based Approach Grid-File indexing Incremental Outlier Approach