A new differential evolution technique, Tree Structure-Based Differential Evolution (TSDE), was proposed by the authors for optimization of trees. However, TSDE supposes the same number of edges in all internal nodes. In this paper, TSDE is improved for problems which need the different number of edges in internal nodes, and the authors show the effectiveness of TSDE by comparing it with conventional methods in simulations. In Evolutionary Computation (EC), the balance of the global search and local search influences the total search performance. Differential Evolution (DE) for optimization in continuous search space implements effective search using difference vectors between two individuals. Tree structure optimization problems can be solved by DE if difference between two trees can be defined. In TSDE, difference between two trees is defined as the set of parts that appear only in one of the two trees, and mutation operations are executed using it. When the two trees are different in their structures, they have a large difference, while two trees with similar structures gives a small difference. TSDE controls a search range using difference between two trees and implements effective search.