Simulation Based Resource Optimization Using a Decision Tree Clearing Function

This study presents a novel approach to resource allocation in software development teams working with Kanban. The simulation algorithm created in this study takes three types of resources, three types of work, resource capabilities, and a blocking mechanism different from the classic machine breakdown scenario. The data generated by the simulations are used to train a decision tree regression which is integrated into an optimization model as a clearing function. In numerical analysis, the research compares the decision tree clearing function to a straightforward two-step model that only takes the best of the simulation data and finds a resource allocation and a greedy heuristic algorithm which starts from an initial feasible solution and improves it step-by-step. Findings show that the developed decision tree clearing function model outperforms the other two benchmark models in mid and high amounts of data.