As web-applications grow in number and complexity, there is a need for efficient mappings from XML schemas to the flat relational tables so that existing functions in relational database systems can be utilized. However, many of the existing mapping methods, such as the model-based or the structure-based methods, do not exploit query history for better query performance. In this paper, we propose the use of genetic algorithm (GA) in a cost-based approach for converting a XML schema to relational tables. By formulating the mapping problem as a cost optimization task with respect to a set of weighted frequent queries, we can obtain an efficient mapping that minimizes the queries execution time. In our experiments, we show that the mapping obtained by GA is superior to other non-cost-based approaches. In particular, the GA approach has out-performed the greedy heuristic in the browsing queries where the accessed attributes are many and scattered.