Search for top-k spatial objects based on various-widths clustering

Search for Top-k spatial objects containing keywords returns the k best objects according to the ranking function which takes both the text relevancy and location proximity into consideration. This is extensively studied and applied among the researches and engineering. However, searching for Top-k objects always leads to high computational cost. This paper proposes an efficient searching method for Top-k spatial objects based on various-widths clustering. The main process of the method can be summarized in four steps: 1) Cluster-Width Learning; 2) Partitioning Process; 3) Merging Process; 4) Searching for Top-k objects according to the ranking function. In addition, we execute extensive experiments that are used to compare with the former study achievements to demonstrate the performance of our research. The algorithm we present reduces the computational cost to some extent, and improves the accuracy of searching which is on the basis of running efficiency.