Research About Directional Nearest Neighbor Query On Uncertain Data

Recently, nearest neighbor query processing based on certain data and uncertain datahas obtained some achievements. However, there is limited work about directional nearestneighbor query on uncertain.Directional nearest neighbor query return the object who is not only the nearestneighbor of querying point, but also in the right direction. Nowadays, the directionalnearest neighbor are: visible nearest neighbor query and directional based surroundernearest neighbor query. Visible nearest neighbor query can be treated as a type of passivedirectional nearest neighbor query; its direction is based on object’s visibility. However,directional based surrounder nearest neighbor query is a type of active direction nearestneighbor query; since its direction is input by the user, and the query return the nearestneighbor of that direction.This paper processes the directional nearest neighbor query on uncertain data andproposes probabilitistic threshold visible nearest neighbor query and directional basedsurrounder nearest neighbor query on uncertain data, besides two algorithms are proposedin this paper to solve those query. First, in the processing of probabilistic threshold visiblebearest neighbor query, the candidate set size is properly proposed by using the definitionof absent probability; and with k means cluter technique, the computing cost canprocessed in batch.Then, in the processing of directional based surrounder nearest neighbor query, byusing effective pruning method which is based on nearest surrounder query, the candaditeset size performances well and at the same time make sure that all candidate objects can becontained in the set. Besides, directional closeness between uncertain objects is properlydefined in this paper. Superseding relationship is chosen as the dominance relationshipbetween two uncertain objects, and with this supersedingship, the result of directionalbased surrounder superseding nearest neighbor query can be returned with onesuperseding nearest neighbor query among all candidate objects.Last, queries proposed in this paper are tested in uniform distribution uncertain database with different parameters, and the experiment analysis is written at the sametime.