Scholarlink

<p>Resource planning plays an important role in matching resources with projects and jobs with multiple attributes. Motivated by real world applications at professional service organizations (PSOs), we coin and study the <i>stochastic resource planning</i> (SRP) problem with both the demand- and supply-side uncertainties due to uncertain bidding outcome and attrition. A two-stage stochastic programming model is developed for the SRP, which is shown to have properties for the model to be transformed to a polynomially solvable minimum cost matching problem. This makes it possible to solve large-scale SRP instances to optimality efficiently. In addition, we develop advanced machine learning techniques to feed the proposed model with the parameters needed, such as the matching score. We present a use case at a PSO for its IT recruiting application. Real life data of resumes and job descriptions from career websites are collected and applied to estimate the matching scores. Comparing with the expected value approach, our SRP solution achieves significantly better performance metrics of staffing cost, resource utilization and job fulfillment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Using optimization and AI for large-scale stochastic resource planning

  • 期刊论文

  • 发表日期：

    2026年7月18日

查看全文
  • Hector G. de Alba,
  • Haitao Li,
  • Andres Tellez-Crespo,
  • Cipriano Arturo Santos,
  • Jose Emmanuel Gomez-Rocha,
  • Marcos C. Vargas,
  • Victor A. Regueira,
  • Lyle Ramshaw,
  • Adrian Ramirez-Nafarrate

摘要

Resource planning plays an important role in matching resources with projects and jobs with multiple attributes. Motivated by real world applications at professional service organizations (PSOs), we coin and study the stochastic resource planning (SRP) problem with both the demand- and supply-side uncertainties due to uncertain bidding outcome and attrition. A two-stage stochastic programming model is developed for the SRP, which is shown to have properties for the model to be transformed to a polynomially solvable minimum cost matching problem. This makes it possible to solve large-scale SRP instances to optimality efficiently. In addition, we develop advanced machine learning techniques to feed the proposed model with the parameters needed, such as the matching score. We present a use case at a PSO for its IT recruiting application. Real life data of resumes and job descriptions from career websites are collected and applied to estimate the matching scores. Comparing with the expected value approach, our SRP solution achieves significantly better performance metrics of staffing cost, resource utilization and job fulfillment.