Recent studies have shown that optical WDM networks with mixed line rate (MLR) support efficiently respond to diverse variety of traffic requirements having different service demands. In MLR WDM optical networks, an intelligent line rate selection strategy can enable cost efficient resource allocations for routing, rate and wavelength assignment problem. In this work, we propose a learning based line rate assignment strategy that reduces communication cost and bandwidth blocking ratio. Performances of three variations of the proposed strategy are compared with maximum and minimum line rate assignment strategies of the literature through simulations. Results show that performance can be improved with the proposed algorithm.