Capital flows are increasingly intertwined globally and, consequently, have brought advantages to global investment strategies. Having a global view of portfolio allocation brings about the diversification of risks in investments. In this paper, a framework to select and optimize asset portfolios in relevant financial markets for short term investment is proposed. In this approach, beta portfolio is a measure of intertwined asset risks and Genetic Relation Algorithm is the evolutionary computing framework for building comprehensible and compact structures of global assets. The algorithm evaluates the relational beta coefficient among assets and generates a robust portfolio in the last generation. Simulations are done using stocks, bonds and currencies as three major asset classes, i.e., the data corresponding to relevant financial markets in USA, Europe and Asia, and the efficiency of the proposed method is compared with traditional Capital Asset Pricing Model(CAPM) for building portfolios. © 2009 SICE.