Ontology serves as a fundamental technique within the Semantic Web by offering a structured framework for knowledge representation across different fields. However, its practical use is hindered by the challenge of ontology heterogeneity, which results from diverse representations of entities. Ontology Matching (OM) addresses this challenge by detecting alignments between entities with identical meanings in various ontologies, using Similarity Features (SFs). However, the complex dependencies and interactions among SFs make manual construction a challenging combinatorial optimization problem, often resulting in premature convergence and reduced alignment accuracy. In response to this issue, this paper proposes an innovative framework designed to autonomously build and adjust advanced SFs to improve the outcomes of matching. In particular, Multi-Objective Genetic Programming (MOGP) is first used to build diverse high-level SFs, while Single-Objective Genetic Programming (SOGP) is then utilized to refine the accuracy and confidence of the final SF. The experimental findings reveal that our approach reliably yields high-quality alignments, offering notable enhancements in both accuracy and efficiency compared to conventional GP-based OM methods.