建筑业劳动力资源的高效配置一直是社会和业界关注的重要问题。但长久以来，农民工作为我国建筑劳务用工的主要力量，却一直游离在规范有形的建筑劳务市场之外，造成供需失衡的问题。对此，政府发布系列政策文件以加快培育建筑产业工人队伍，实现工人就业高效有序的目标。 依靠建筑产业互联网平台等信息技术，海量建筑劳务信息得以流通，为供需匹配提供条件。 然而现有研究主要关注工程项目效益最大化，忽视了以建筑工人为主的供给端需求，缺乏供需匹配的高效方法，无法实现供需失衡的有效调节。因此，综合考虑供需双方需求偏好，利用强化学习改进遗传算法（RL-GA）求解供需匹配优化问题，对促进建筑劳动力资源高效配置具有重要意义。
本文剖析了我国建筑劳务用工模式的发展，总结专业作业企业用工模式的优势特点，针对该模式下供需匹配的独特性，提出通过供需匹配度衡量供需双方匹配偏好的方法，并依此建立建筑劳务服务供需匹配优化问题的整数规划模型。 根据模型分析适用方法，引入遗传算法对问题模型优化求解，基于问题的一个实例设计两轮求解实验，验证算法的可行性，并在追求算法更高计算效率和更优求解质量的方向上探索算法改进的可能性。 为提高算法性能， 引入强化学习改进遗传算法，构建问题模型优化求解的 RL-GA 算法，通过对同一个问题实例进行实验，并对比单一遗传算法和 RLGA 算法在求解结果上的表现，证明了采用 RL-GA 算法求解建筑劳务服务供需匹配优化问题具有的优势，并将 RL-GA 算法在测试函数集上与其他算法进行测试对比，说明该求解方法具有较广的适用性与较好的应用价值。
研究结果表明，专业作业企业用工模式更适合建筑产业工人队伍建设的需要，在该用工模式下进行建筑劳务服务供需匹配，采用单一遗传算法具有问题优化求解的可行性，但是算法计算效率和求解效果无法同时得到保证，采用 RL-GA 算法能在单一遗传算法的基础上，全方位改进问题求解的性能表现，实现建筑业劳动力资源的高效配置，促进建筑产业工人队伍的培育。

关键词：建筑工人产业化；建筑劳务；供需匹配；专业作业企业；强化学习改进遗传算法（RL-GA）

The efficient allocation of labor resources in the construction industry has always been an important issue of concern to society and the industry. However, for a long time, as the main force of China's construction workforce, migrant workers have been dissociated from the normative and legal construction labor market, resulting in an imbalance between supply and demand. In this regard, the government has issued a series of policy documents to accelerate the cultivation of construction industrial workers and achieve the goal of efficient and orderly employment of workers. Relying on information technology such as the construction industry internet platform, massive construction labor service information can be circulated, providing conditions for supply and demand matching. However, existing research mainly focuses on maximizing the benefits of construction projects and ignores the supply-side demand dominated by construction workers. It lacks an efficient method of matching supply and demand and cannot effectively adjust the imbalance between supply and demand. Therefore, comprehensively considering the demand preferences of both supply and demand sides, it is of great significance to promote the efficient allocation of construction labor resources by using reinforcement learning to improve genetic algorithm (RL-GA) to solve the optimization problem of supply and demand matching.
This research analyzes the development of China's construction labor employment mode and summarizes the characteristics and advantages of the employment mode of professional construction enterprises. Aiming at its uniqueness in the matching of supply and demand, a method of measuring the matching preference of supply and demand sides through the matching degree of supply and demand is proposed, and an integer programming model for the matching optimization of supply and demand of construction labor services is established. According to the model, the applicable method is analyzed, and the genetic algorithm is introduced to optimize the problem model. Two rounds of solution experiments are designed to verify the feasibility of the algorithm through an instance of the problem and explore possible directions for algorithm improvement in the pursuit of higher computational efficiency and better solution quality. To improve algorithm performance, reinforcement learning is introduced to improve the genetic algorithm, and an RL-GA algorithm is constructed to optimize the problem model. Through the experiment on the same instance of the problem, and comparing the performance of the single genetic algorithm and the RL-GA algorithm in the solution results, it is proved that the RL-GA algorithm has an advantage in solving the matching optimization problem of supply and demand for construction labor services. And the RL-GA algorithm is tested and compared with other algorithms on the CEC 2017 benchmark test functions, which shows that the solution method has wider applicability and better application value.
The research results show that the employment mode of professional construction enterprises is more suitable for the needs of the cultivation of construction industrial workers. Matching the supply and demand of construction labor services under this employment model, it is found feasible to use a single genetic algorithm to optimize the solution of the problem, but the algorithm computational efficiency and solution quality cannot be guaranteed at the same time. The RL-GA algorithm can improve the performance of problem solving comprehensively on the basis of a single genetic algorithm, realize efficient allocation of labor resources in the construction industry, and promote the cultivation of construction industrial workers.