In this paper, 62 cases of tall buildings with transfer story are collected, a unified expression mode of case information in the building is set up for structure scheme design. In choosing structural style for a tall building with transfer story, main control factors are drawn, with which a mathematical model is formed on the basis of BP (back-propagation) neural networks. Traditional BP algorithm and levernberg-Marquert (L-M) algorithm are adopted to research the style-selection of transfer story in tall buildings, respectively. Indicated by the research results, L-M algorithm solves the style-selection preferably for the complicated tall building structures with transfer story. This paper also paves the way for further development of the integrated, intelligent, scientific stle-selection of complicated tall buildings with transfer story.