Cost-Effective Viral Marketing in the Latency Aware Independent Cascade Model

A time-constrained viral marketing campaign allows a business to promote a product or event to social network users within a certain time duration. To perform a time-constrained campaign, existing works select the duration of the campaign, and then a set of k seeds that maximize the spread (expected number of users to which the product or event is promoted) for the selected duration. In practice, however, there are many alternative durations, which determine the monetary cost of the campaign and lead to seeds with substantially different spread. In this work, we aim to select the duration of the campaign and a set of k seeds, so that the campaign has the maximum spread-to-cost ratio (i.e., cost-effectiveness). We formulate this task as an optimization problem, under the LAIC information diffusion model. The problem is challenging to solve efficiently, particularly when there are many alternative durations. Thus, we develop an approximation algorithm that employs dynamic programming to compute the spread of seeds for several possible durations simultaneously. We also introduce a new optimization technique that is able to provide an additional performance speed-up by pruning durations that cannot lead to a solution. Experiments on real and synthetic data show the effectiveness and efficiency of our algorithm.