Smart meters networks are rapidly becoming a reality in many developed countries. In this paper, we focus on the optimization of a network of smart meters operated by the wM-Bus protocol, which is the de-facto metering standard in Europe. In such a scenario, data concentrators receive data from smart meters using the wM-Bus protocol and relay it to a central server using a legacy mobile cellular backhauling technology such as GSM/GPRS. Due to the massive amount of data produced by meters installed in urban scenarios and the association-less nature of the wM-Bus protocol, data concentrators may be overloaded with many duplicate measurement packets, causing capacity problems on the backhauling links and computational overload at the central server. To solve these issues, we propose a datadriven optimization framework to populate forwarding whitelists at each data concentrator so that (i) load is balanced among the different concentrators and (ii) the overall performance of the network in terms of packet reception rate and received signal strength are maximized. We also propose a heuristic algorithm to generate near optimal forwarding whitelists in acceptable computing time. Extensive experiments are performed on a real scenario consisting of a city wide gas meter network deployed in northern Italy. Results show that the proposed heuristic is able to produce whitelists that reduce the average backhauling traffic as much as 80%, with a corresponding network quality within 4% of the one computed by the optimal solution.