This paper proposes a binary classification tree aiming at solving multi-class classification problems using binary classifiers. The tree design is achieved in a way that a class group is partitioned into two distinct subgroups at a node. The node adopts the class-modular scheme to improve the binary classification capability. The partitioning is formulated as an optimization problem and a genetic algorithm is proposed to solve the optimization problem. The binary classification tree is compared to the conventional methods in terms of classification accuracy and timing efficiency. Experiments were performed with numeral recognition and touching-numeral pair recognition.