Predictive machine learning for prescriptive applications: A coupled training–validating approach

作者：

Highlights：

• Proposed coupled training-validating approach to train predictive learning models

• Method considers prescription loss as the objective for hyper-parameter calibration.

• The coupled approach is applicable to most predictive machine learning models.

• Experiments with synthetic and real data demonstrate promising results.

摘要

•Proposed coupled training-validating approach to train predictive learning models•Method considers prescription loss as the objective for hyper-parameter calibration.•The coupled approach is applicable to most predictive machine learning models.•Experiments with synthetic and real data demonstrate promising results.

论文关键词：Predictive modeling,Deterministic optimization,Machine learning,Stochastic optimization,Knowledge discovery

论文评审过程：Received 19 October 2021, Revised 14 May 2022, Accepted 16 May 2022, Available online 25 May 2022, Version of Record 4 June 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109080