Optimization of architectural design scheme based on artificial intelligence algorithm

Aiming at the problems of inefficiency, information disjunction and lack of innovation in traditional architectural design schemes, this paper proposes an intelligent evaluation model combining grey correlation analysis, radial basis function neural network and ant colony algorithm. By constructing multi-dimensional building performance index system, combining data dimensionality reduction and nonlinear mapping technology, the problem of index homogeneity and parameter convergence in traditional evaluation is solved. Through experiments, it is found that the average evaluation accuracy of ACO-RBF and GRA-RBF models is 87.48% and 86.60% respectively, while the fusion model in this paper achieves a performance improvement with an accuracy of 93.31%. The model adopts a three-stage collaborative mechanism -- feature space reconstruction, parameter global optimization and nonlinear decision mapping. It significantly improves the objectivity and reliability of complex building performance evaluation, provides a verifiable technical framework for intelligence-driven design decisions, and promotes the transformation of architectural design scheme evaluation from an experience-dependent paradigm to a data-driven paradigm.