This paper proposes a novel data-driven optimization framework that integrates building design decisions with simulation results to optimize sustainability across environmental, social, and economic dimensions. The building industry faces increasing pressure to reduce emissions and environmental impacts, but existing design tools do not fully account for how buildings will be operated by occupants. The proposed optimization framework utilizes statistical learning via metamodels and decision optimization to explore the complex tradeoffs involved in building decisions. The focus of this paper is on developing an optimization approach performed on common machine learning models such as classification and regression trees (CART), gradient boosted trees and random forest to investigate the causal relationship between building decisions and performance metrics. We integrate both descriptive and prescriptive analytics to inform an optimization model that provides decisions for both the design and operations of a given building. A case study is presented demonstrating the potential benefits of the optimization model in the design phase of a building. The integrative, data-driven optimization approach incorporates both categorical and continuous variables to create higher-performing, more sustainable building designs that better balance Planet, People, and Prosperity goals.