Optimizing resource redistribution in delayed projects using genetic algorithms

Abstract

Efficient resource management is crucial in project execution, especially in the construction industry, where delays are common and significantly impact costs and quality. This study focuses on evaluating the effectiveness of a genetic algorithm for optimal resource redistribution in projects with delays using an experimental design. Independent variables were manipulated based on genetic algorithm parameters to observe their effects on resource redistribution effectiveness and efficiency. This study population consisted of simulated resource planning activities, representative of construction industry environments, with a selected sample of 10 activities each lasting 44 working days. Initial distribution was generated using the normal distribution, with a mean (μ) of 22 and standard deviation (σ) of 8.8. To simulate delays, values for the first 10 days were adjusted to reflect a cumulative delay of 10%. To evaluate the robustness and effectiveness of the genetic algorithm, multiple instances of activities with different initial distributions were generated, simulating various scenarios. Python was used for algorithm implementation, employing techniques such as tournament selection, single-point crossover, and uniform mutation. The fitness function, based on the Mean Squared Error (MSE), measured the quality of resource redistribution in terms of minimizing cumulative delay and resource efficiency.