Contextual Optimizer through Neighborhood Estimation for prescriptive analysis

TL;DR

This paper introduces CONE, a novel approach combining neighborhood estimation and sampling to improve contextual decision-making under heteroscedastic noise, validated through theoretical analysis, benchmarking, and real-world application.

Contribution

It presents a new consistent neighborhood estimation method and a rate-efficient sampling rule, advancing prescriptive analysis in noisy, real-world environments.

Findings

Theoretical validation of CONE's effectiveness.

Improved performance in numerical benchmarks.

Successful deployment in hospital staffing optimization.

Abstract

We address the challenges posed by heteroscedastic noise in contextual decision-making. We propose a consistent Shrinking Neighborhood Estimation (SNE) technique that successfully estimates contextual performance under unpredictable variances. Furthermore, we propose a Rate-Efficient Sampling rule designed to enhance the performance of the SNE. The effectiveness of the combined solution ``Contextual Optimizer through Neighborhood Estimation"(CONE) is validated through theorems and numerical benchmarking. The methodologies have been further deployed to address a staffing challenge in a hospital call center, exemplifying their substantial impact and practical utility in real-world scenarios.