Spatio-Temporal Outlier Detection Technique

Outlier detection is very important functionality of data mining, it has enormous applications. This paper proposes a clustering based approach for outlier detection using spatio-temporal data. It uses three step approach to detect spatiotemporal outliers. In the first step of outlier detection, clustering is performed on the spatio-temporal dataset with proposed Spatio-Temporal Shared Nearest Neighbor (ST-SNN) clustering approach, which is capable to handle high dimensional spatio-temporal data having different sizes and densities and also capable to identify arbitrary shaped cluster. Proposed clustering approach first finds nearest neighbors of each data points and after that it finds the shared nearest neighbor similarity between pair of points in terms of how many nearest neighbors the two points share. Using this similarity measure, our algorithm identifies core points and build clusters around the core points. In the second step of outlier detection, spatial outliers are identified. Finally, in the third step, to find presence of outliers in our dataset, identified spatial outliers are compared with temporal neighbors. The experimental results show that proposed approach is performing much better to identify outliers, especially in high dimensional spatial-temporal data.