Enterprise Wireless LANs (E-WLANs) such as airport WiFi, have become a convenient way for Internet access for mobile users. In an E-WLAN, access points (APs) are usually deployed with high-density around the infrastructure to provide sufficient coverage and for a better service, where a mobile user chooses one AP to associate with among multiple available APs in the vicinity. Many studies have been done on developing user association techniques to increase system performance, with various objectives including network throughput maximization, load balancing etc. Our work is unique in that we focused on bandwidth cost minimization via user association from the perspective of the E-WLAN operators. Specifically, by considering the bandwidth demands from mobile users, we modeled the joint user association and cost minimization problem in the heterogeneous E-WLAN with additional constraints from individual bandwidth demands as an optimization problem. To solve the optimization problem efficiently, we propose an approximation algorithm using relaxation and rounding techniques. We prove that the proposed algorithm has performance bound with a constant ratio to the optimization problem. Furthermore, our simulation results exhibit the superiority of our proposed algorithm over prior schemes.