Estimation of distribution algorithms (EDAs) are effective methods of solving complex combinatorial problems. This paper explores an enhanced hybrid approach to EDAs for solving complex combinatorial optimization tasks, focusing on the Traveling Salesman Problem (TSP). Our method introduces the concept of overlapped subpopulations, inspired by the island model of evolution, to simulate genetic diversity and migration across subdivisions within the population. This adaptation allows each subpopulation to independently generate probabilistic models, fostering localized optimization while promoting cross-subpopulation learning. Experimental results demonstrate that this method achieves higher solution quality than traditional EDAs without population subdivisions, using selected discrete and continuous, acclaimed benchmarks. These findings suggest that overlapping subpopulations enhance EDA performance by balancing global and local search capabilities, offering a robust alternative for complex optimization scenarios.