\(\text {SPL}^{{index}}\) : A Spatial Polygon Learned Index

Effective indexing is crucial for any AI and big data analysis task involving huge datasets. Recently, machine-learned models for indexing have achieved much attention, and we apply such for spatial data, specifically huge collections of polygons. We propose an index structure \(\text {SPL}^{{index}}\) that organizes polygons into a tree of clusters with linear regression models for effective branching in search. It integrates an effective layout of polygon data to disk space that minimize disk access and amount of data to be kept in main memory. The approach is shown outperform the state-of-the-art R-tree for both range and point queries.