Deep trajectory: a deep learning approach for mobile advertising in vehicular networks

Abstract

Mobile advertising is an interesting research topic and a promising commercial application in vehicular networks, which is expected to spread the timely information into the network at a lower cost. Seed vehicles are of vital importance for information broadcasting in mobile advertising. How to find the best set of seed vehicles with maximal influence on the network is a very challenging problem. In this paper, we propose developing deep learning schemes to select the most suitable candidate vehicles for mobile advertising so as to achieve maximum coverage and impact in vehicular networks. Specifically, we adopt the deep structure for the mobile advertising problem to exploit the spatial-temporal regularities of vehicle traces. The adaptive models MLP-W and MLP-T are able to either infer the future topology of vehicles explicitly or predict the future trajectory and infer topology implicitly. Then, vehicles with higher centralities are considered as good seeds for timely information disseminating. Extensive experiments have been conducted over two real-world taxi trace datasets. And the experimental results demonstrate the efficacy of our proposed methodology compared with the state-of-the-art approach for mobile advertising tasks.