Nearest neighbor imputation algorithms: a critical evaluation.

Beretta L, Santaniello A. · 2016

Read Nearest neighbor imputation algorithms: a critical evaluation. by Beretta L, Santaniello A. (2016) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4959387.txt.

Description

Lorenzo Beretta and Alessandro Santaniello test nearest neighbor methods for filling gaps in datasets. Their study appeared in BMC Medical Informatics and Decision Making in 2016. Missing values are common in medical data. Imputation replaces each gap with a plausible value borrowed from similar records. A good method should stay accurate without warping the shape of the data. The team ran simulations with set patterns and amounts of missingness, under several learning setups.

Using several neighbors beat using just one on precision in nearly every setup. The single neighbor method, though, was the only one that left the data structure intact. Large neighbor counts looked precise yet badly distorted the true distribution. Adding a feature filter called ReliefF, with a low neighbor count, gave the best trade off. The authors warn that imputation choices quietly shape later statistics, so the method must be picked with the analysis goal in mind.

Catalog record

CREATORBeretta L, Santaniello A.
YEAR2016
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC4959387

Browse the full library · About the Tower of Babel