In heterogeneous scenarios, nodes greatly differ with respect to their capabilities and mobility patterns. Moreover, episodic connectivity in opportunistic networks further aggravates the problem of finding a suitable next-hop to obviate unnecessary utilization of network resources. In this paper, we present a Multiattribute Routing Scheme (MARS) based on “Simple Multiattribute Rating Technique” (SMART) that collects samples of vital information about a node’s different characteristics. This stochastic picture of a node behavior in multiple dimensions is then effectively employed in calculating its next-hop fitness. We also devise a method based on learning rules of neural networks which dynamically determines relative importance of each dimension to maximize next-hop utility of a node. With simulations, using synthetic and real mobility traces against well-known utility-based schemes, we show that MARS can achieve better delivery ratios despite introducing limited redundancy within the network. [ABSTRACT FROM AUTHOR]