A Generic Framework for Handling Uncertain Data with Local Correlations

Data uncertainty is ubiquitous in many real-world applications suchas sensor/RFID data analysis. In this paper, we investigate uncer-tain data that exhibit local correlations, that is, each uncertain ob-ject is only locally correlated with a small subset of data, whilebeing independent of others. We propose a generic framework fordealing with this kind of uncertain and locally correlated data, inwhich we investigate a classical spatial query, nearest neighborquery, on uncertain data with local correlations (namely LC-PNN).Most importantly, to enable fast LC-PNN query processing, wepropose a novel filtering technique via offline pre-computations toreduce the query search space. We demonstrate through extensiveexperiments the efficiency and effectiveness of our approaches.