A data-driven meta-learning recommendation model for multi-mode resource constrained project scheduling problem

Xianghua Chu

doi:10.1016/j.cor.2023.106290

A data-driven meta-learning recommendation model for multi-mode resource constrained project scheduling problem

2023-09-01

7

PRE

AI

DOI：10.1016/j.cor.2023.106290

Original

Original request for help

Save

Abstract

En 中文

Meta-heuristics widely proposed in addressing multi-mode resource constrained project scheduling problem (MRCPSP) are problem-dependent. This paper first proposes an adaptive data-driven meta-learning Meta-heuristic Recommendation Model (MRM) to solve MRCPSP intelligently and efficiently. By learning the asso-ciation between problem meta-features and algorithm performance, MRM can identify the most appropriate algorithm for different MRCPSPs. Multiclass Support Vector Machine (MCSVM) are integrated to train the classifiers for predicting the performance of the candidate meta-heuristics. To validate the proposed MRM, the performance is evaluated and compared in terms of accuracy, precision, sensitivity, and comprehensive evalu-ation index. In the experiments of 4 scenarios with 2 strategies, the average optimization and prediction accu-racies are higher than 90% without increase in computational complexity. Comprehensive experiments and numerical results demonstrate the outperforming performance of the proposed MRM across various MRCPSP.

Keywords:

Data -driven decision making

Meta-learning

Feature extraction

Meta-heuristics

A data-driven meta-learning recommendation model for multi-mode resource constrained project scheduling problem

Abstract

Journal

Organization