In projects, resource leveling is done to reach lower resource fluctuations in order to avoid the problem of sudden resource needs or excessive resource accumulation. The biggest challenge in doing resource leveling is the increasing project duration. In addition, traditional analytical and heuristic approaches when solving construction resource leveling problems are inefficient and inflexible. Based on this, in order to effectively provide the good quality combination of multiple construction resources, a genetic algorithm model was employed in this study to overcome the drawbacks of traditional construction resource leveling algorithms. The project contains activities interrelated by finish-start type precedence relations with a zero-time lag, which require a set of renewable resources. Delay amounts of activities are determined by a genetic algorithm to level resources and minimize project time. Model validations are performed to demonstrate the model's capability and applications. The modified genetic algorithm was tested on benchmark construction resource-leveling instances and compared with standard genetic algorithm, particle swarm optimization, and ant colony optimization. Under the tested conditions, it achieved performance improvements of 49%, 55%, and 43%, respectively, and converged faster by 44%, 49%, and 59%. It was concluded that the applied revisions and modification of the genetic algorithm procedure, such as the usage of the dynamic probability in creating the next generations, could possibly improve the rapidity and the efficiency of the upgrade process and enhance its ability to arrive at quick solutions.