An Algorithm for Generating a Composite Proxy Target Variable for Business Decision-related Machine Learning Models

Business decision-making methods often lack information about a target or outcome variable. In this study, we propose a multi-criteria decision analysis (MCDA)-based composite proxy target variable generation technique for business decision modeling that uses relevant features or independent variables conceptually related to the intended target variable. Using synthetic brick-and-mortar store expansion data and scenarios, we found that MCDA-based composite proxy variables provide a second-best approach for generating a proxy target variable when the required data are either not available or limited.

#Paper link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4277180