Features are fundamental components of decision tree modeling, and their relevance, quality, and selection are crucial determinants of the model's effectiveness and performance. However, decision trees can be computationally expensive, requiring a significant amount of memory to store the trees and their associated data structures. To address this limitation, we present a novel approach that utilizes a Cuckoo Search-based feature selection algorithm to construct efficient and optimal decision trees. The Cuckoo Search algorithm, inspired by the behavior of cuckoo birds, is a powerful metaheuristic algorithm that effectively selects high-quality features and creates accurate decision trees in the subforest. We evaluate the proposed method on a variety of datasets from the standard UCI learning repository with different domains and sizes, and our results demonstrate that the algorithm creates optimal decision trees with high performance.