A stochastic approximation approach to simultaneous feature weighting and selection for nearest neighbour learners

作者：

Highlights：

• We introduce a new feature selection and weighting methodology for nearest neighbour learners.

• This methodology is based on simultaneous perturbation stochastic approximation.

• We present computational experiments involving both regression and classification problems.

• Our methodology outperforms several other state-of-the-art feature weighting algorithms.

摘要

•We introduce a new feature selection and weighting methodology for nearest neighbour learners.•This methodology is based on simultaneous perturbation stochastic approximation.•We present computational experiments involving both regression and classification problems.•Our methodology outperforms several other state-of-the-art feature weighting algorithms.

论文关键词：Nearest neighbour learner,Feature weighting,Feature selection,Stochastic approximation,Gradient descent optimisation

论文评审过程：Received 25 June 2020, Revised 10 September 2020, Accepted 24 July 2021, Available online 29 July 2021, Version of Record 4 August 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115671