Fast and Efficient Approach For Nearest Neighbor Search

Conventional spatial queries, whereas range search and nearest neighbor retrieval, involve only conditions on the objects geometric properties. Many modern applications call for the novel forms of queries that aim to find objects which are satisfying both the spatial predicate, and the predicate on the their associated texts. For example, instead of searching all the restaurants, nearest neighbor query would instead ask for restaurant that is the closest among them whose menus contain steak, spaghetti, brandy all at same time. The best solution given to such queries is based on the IR2-tree, which shown in this paper, has few deficiencies that seriously impact on the efficiency. By this, we develop the new access method known as spatial inverted index that will extends the conventional inverted index which cope with multidimensional data, where it comes with algorithms that can be answer nearest neighbor queries with search keywords in the real time. As verified by an experiment, the proposed techniques perform the IR2-tree in the query response time significantly, often by the factor of orders of magnitude.