Scholarlink

<p>In this study, we consider that the resource transfer time and activity duration are stochastic variables, and new projects arrive randomly during the multi-project execution. The performance of 72 priority rules (PRs) is evaluated for the resource-constrained multi-project scheduling problem under dynamic and uncertain environments (RCMPSP-DUE), which combines 18 activity PRs with 4 resource transfer PRs. Among the 18 activity PRs, 8 activity PRs include 4 newly designed activity PRs based on the available stochastic resource transfer information and 4 composite activity PRs that consider activity attributes and resource transfer information simultaneously. The performance of PRs is evaluated under different objectives from the project, multi-project and resource perspectives. In addition, the machine learning based decision tree algorithm is designed for adaptive recommendation rules based on project features. Extensive numerical experiments show that the classical activity PRs for the multi-project scheduling problem do not perform best for the RCMPSP-DUE. The decision tree algorithm we designed can more directly and efficiently recommend rules based on project characteristics. We can suggest higher-performing PRs tailored to various objects and uncertain environments. The recommendation performance of J48 and random forest decision tree algorithm is better, and when multiple indicators of projects are used simultaneously, the performance of the decision tree model reaches its better state.</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

Efficient priority rules for the resource-constrained multi-project scheduling problem under dynamic and uncertain environments

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Lubo Li,
  • Haohua Zhang,
  • Erik Demeulemeester,
  • Sijun Bai

摘要

In this study, we consider that the resource transfer time and activity duration are stochastic variables, and new projects arrive randomly during the multi-project execution. The performance of 72 priority rules (PRs) is evaluated for the resource-constrained multi-project scheduling problem under dynamic and uncertain environments (RCMPSP-DUE), which combines 18 activity PRs with 4 resource transfer PRs. Among the 18 activity PRs, 8 activity PRs include 4 newly designed activity PRs based on the available stochastic resource transfer information and 4 composite activity PRs that consider activity attributes and resource transfer information simultaneously. The performance of PRs is evaluated under different objectives from the project, multi-project and resource perspectives. In addition, the machine learning based decision tree algorithm is designed for adaptive recommendation rules based on project features. Extensive numerical experiments show that the classical activity PRs for the multi-project scheduling problem do not perform best for the RCMPSP-DUE. The decision tree algorithm we designed can more directly and efficiently recommend rules based on project characteristics. We can suggest higher-performing PRs tailored to various objects and uncertain environments. The recommendation performance of J48 and random forest decision tree algorithm is better, and when multiple indicators of projects are used simultaneously, the performance of the decision tree model reaches its better state.