计算机工程与应用2026，Vol.62Issue(5)：346-363,18.DOI:10.3778/j.issn.1002-8331.2411-0010
数据驱动的资源受限项目调度问题求解器推荐研究
Data-Driven Research on Solver Recommendation for Resource-Constrained Project Scheduling Problems
摘要
Abstract
The resource constrained project scheduling problem(RCPSP)is widely used in engineering management and other fields,and it is very important to solve this problem efficiently for project management.However,due to the inherent NP-hard characteristics of RCPSP,the performance of the existing solution methods shows a strong dependence on project instances,and it is difficult to find a general and efficient algorithm.To this end,a data-driven RCPSP solver recommenda-tion framework is proposed to realize intelligent algorithm selection for different project examples,so as to overcome the blindness of existing algorithm selection schemes and improve the solution efficiency.The construction of this framework stems from the insight into the complex relationship between RCPSP problem characteristics and algorithm performance,and attempts to use machine learning methods to mine this latent relationship and transform it into knowledge to guide algo-rithm selection.Firstly,the RCPSP solution algorithm recommendation dataset is constructed,which includes three-dimen-sional feature sets:network topology,resource and time.Then,the feature selection method is combined to extract the opti-mal feature subset,and a recommendation model based on tree ensemble algorithm is constructed to learn the internal laws of this complex mapping relationship and achieve accurate algorithm recommendation.Finally,the SHAP model is used to analyze the attribution analysis of the recommendation model,analyze the key project characteristics that affect the algorithm selection,and provide more explanatory decision support for project managers.Experimental results show that the proposed recommendation framework has a recommendation accuracy of more than 70％on the four datasets,and is better than other recommendation algorithms in various indicators.Characteristics such as resource intensity,lower bound of project duration,and network width are proved to have an important impact on algorithm selection.This study verifies the feasibility and effectiveness of the data-driven method in solving the problem of RCPSP algorithm selection,provides a scientific and intelligent algorithm selection scheme for project managers,effectively reduces the difficulty of decision-making,and helps to improve the efficiency of project management.关键词
资源受限项目调度/求解器推荐/数据驱动/树集成算法/SHAP模型Key words
resource-constrained project scheduling/solver recommendation/data-driven/tree ensemble algorithms/SHAP model分类
信息技术与安全科学引用本文复制引用
曾鸣,戴业东,刘万安..数据驱动的资源受限项目调度问题求解器推荐研究[J].计算机工程与应用,2026,62(5):346-363,18.基金项目
国家自然科学基金(72401079). （72401079）
