Parameter-free surrounding neighborhood based regression methods

作者：

Highlights：

• Parameter-free surrounding neighborhood (PSN) is introduced for NN regression.

• PSNs are based on proximity graphs, i.e. MST, RNG and GG.

• PSNs yield a unique neighborhood for each point.

• PSNs consider distance, connectivity and geometrical positions for neighborhood.

• Statistical tests show the superior performances of PSN regression methods.

摘要

•Parameter-free surrounding neighborhood (PSN) is introduced for NN regression.•PSNs are based on proximity graphs, i.e. MST, RNG and GG.•PSNs yield a unique neighborhood for each point.•PSNs consider distance, connectivity and geometrical positions for neighborhood.•Statistical tests show the superior performances of PSN regression methods.

论文关键词：Prediction,k-nearest regression,Minimum spanning tree,Relative neighborhood graph,Gabriel graph

论文评审过程：Received 16 July 2021, Revised 15 October 2021, Accepted 10 March 2022, Available online 19 March 2022, Version of Record 28 March 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116881