In this paper, we present a study of asset allocation using genetic algorithms. This method extends a previous version of a stock selection model using Genetic Algorithms (GA) for solving the problem of asset allocation. The GA is used for optimization of model parameters, feature selection as well as the construction of the Pareto front. On top of that, we proposed another GA to search for the optimal allocation of assets. We then present an investigation for this line of research using financial data of various industrial sectors in Taiwan's stock market. Our experimental results show that our proposed method is capable of significantly outperforming the benchmark and the canonical Markowitz method for asset allocation. Based upon these promising results, we expect this GA methodology to advance the current state of research in soft computing for the real-world applications in asset-allocation.