ZFT INDEX:LEARNING MULTI-DIMENSIONAL INDEX BASED ON PIECEWISE LINEAR REGRESSION
The traditional indexing method is generally a general data structure,which is not designed or optimized for data distribution and characteristics.With the increase of data space dimension or data volume,it may lead to large storage consumption and sharp decline of query efficiency.To solve these problems,this paper proposes a ZFT index(Z-order fitting tree index),which is mainly divided into offline and online parts.The offline construction part used Z-order curve to map data points in multidimensional space to one-dimensional space,and constructed a linear regression model to learn the distribution and characteristics of the mapped data.The online part completed the point query and range query.The experimental results show that the spatial efficiency and query efficiency of ZFT index are significantly better than those of traditional R tree and UB tree,and the speed of range query and model training is better than that of ZM index.
万方数据
维普
